Breaking
Advertisement

AI

AI

Block Unveils Dessa, a Full-Duplex Speech Model Trained Only on Speech

Block’s AI research lab has released Dessa, a full-duplex speech model trained only on human conversation. Block announced it on October 23, 2026 as a research preview. Dessa is a speech-to-speech model from Block that listens and talks at the same time. Block says it used no text and no pretrained models. That is a different design from the text-first stacks behind most voice agents today. TL;DR Runs on: A MacBook Pro M4 Max or better, locally, faster than real time. Performance: Demo-only so far. Live recordings show backchannels, mid-sentence laughs and natural interruptions. Best: Trained on 1 NVIDIA GB200 NVL72 rack (72 GPUs). Moshi used 1,016 H100 GPUs. Worst: 0 published benchmark scores. No weights, paper, API or price. What is Block’s Dessa? Dessa is a true speech-to-speech model: audio in, audio out. According to Block’s project page, everything it knows came from human speech. It is full-duplex, so it listens while it talks. Block says it learned from real conversation when to cut in. It also learned when to backchannel with short cues like “yeah” or “got it”. Block’s demo clips show Dessa laughing mid-sentence and getting louder when asked to speak up. One clip shows it talking with OpenAI’s GPT-Live-1. Block says nobody steered any clip. How does speech-only training work? Most voice systems chain 3 parts: speech recognition, a text model, then speech synthesis. Block team argues that chain throws away emotion, sarcasm, laughter and sighs. Dessa hears and produces those signals directly. Block has not published the architecture. Here is what it has disclosed: Training data: human conversation audio only. No text and no pretrained models. Compute: 1 NVIDIA GB200 NVL72, which links 72 Blackwell GPUs and 36 Grace CPUs. Data efficiency: Block says Dessa learned language from hundreds of times fewer words than LLMs need. Inference: faster than real time, on a MacBook Pro M4 Max or better. The lab also explains why this is hard. Speech never repeats exactly. A spoken word takes thousands of times more bytes than the same word in text. Kyutai’s Moshi, the best-known open full-duplex model, takes another route. Moshi starts from a text language model backbone and predicts time-aligned text tokens before audio tokens. Its model card lists 7 million hours of pretraining audio and 1,016 H100 GPUs. (function(){var f=document.getElementById('mtp-dessa-frame'); window.addEventListener('message',function(e){if(e.source!==f.contentWindow)return;var d=e.data;if(d&&d.mtpDessaH){f.style.height=d.mtpDessaH+'px';}});})(); How does Dessa compare with Moshi, PersonaPlex and GPT-Live-1? Key Takeaways Dessa is full-duplex and speech-to-speech, trained only on human conversation audio. Block trained it on 1 GB200 NVL72 rack; Moshi used 1,016 H100s. It runs locally on a MacBook Pro M4 Max, faster than real time. Block published 0 benchmarks, no parameter count and no weights. Access is a waitlist. Check out the Dessa project page. Compare it with the Moshi paper, the PersonaPlex model card and the GPT-Live-1 docs. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Block Unveils Dessa, a Full-Duplex Speech Model Trained Only on Speech appeared first on MarkTechPost.

AI

AssemblyAI Ships AssemblyAI-TTS-1: A $15 per 1M Character Text-to-Speech API for Voice Agents

AssemblyAI has released AssemblyAI-TTS-1, its first text-to-speech model, now generally available. The AssemblyAI-TTS-1 text-to-speech API targets voice agents running on live phone calls. In that setting, latency and pronunciation decide whether a caller stays on the line. Pricing is $15 per 1M characters, with no annual contract and no concurrency tier to buy. For developers already using AssemblyAI speech-to-text, it closes the voice loop with one vendor. TL;DR Size: It is a hosted API model. Runs on: AssemblyAI’s cloud only, in the US (us-east-1) or EU (eu-north-1). No open weights or on-prem option. Any client with HTTP or WebSocket works. Performance: Ties ElevenLabs v4 on entity accuracy in AssemblyAI’s tests, at under 1/5 of its price. Best: 47 ms median time to first byte across 6 models in AssemblyAI’s API test (self-reported). Worst: 159 ms median streaming latency on Cekura’s independent benchmark, 8th of 19 timed models. Bottom line: Best: $15 per 1M characters with unlimited concurrency. Worst: no SSML, speed control, pronunciation dictionaries or emotion tags yet. What is AssemblyAI-TTS-1? AssemblyAI-TTS-1 is a streaming text-to-speech model from AssemblyAI that turns agent replies into speech for live calls. It was trained on millions of hours of audio, including real conversational data. Most TTS models represent audio as discrete tokens. AssemblyAI says TTS-1 operates on continuous audio representations instead. The AssemblyAI claims this improves realism and cuts time to first spoken audio. The launch includes a library of 4,000+ voices and instant voice cloning. A clone needs a short recording and a confirmation of the speaker’s permission. The clone in the launch post used 24 seconds of audio. Six languages are supported at launch: English, Spanish, French, German, Italian and Portuguese. How do developers integrate AssemblyAI-TTS-1? There are 3 interfaces: an HTTP API, a WebSocket API and an OpenAI-compatible route. The WebSocket session keeps one connection per conversation and accepts a cancel message for barge-in. Output is mono PCM, mu-law or A-law, at 8 kHz to 48 kHz. 24 kHz PCM is the native format; telephony stacks can request 8 kHz mu-law directly. Word timestamps are included. AssemblyAI also published LiveKit and Pipecat reference agents. How does AssemblyAI-TTS-1 perform on benchmarks? AssemblyAI tested TTS-1 against 5 models on 4 English sets, using 50 matched cloned voices. All figures below are self-reported, from an eval dated October 28, 2026. Entity accuracy (580 hard targets such as phone numbers and codes): 93.7%, vs ElevenLabs v4 93.6%, Cartesia Sonic 3.6 82.5%, ElevenLabs Flash v2.5 68.0%. Word accuracy: 99.0% on Seed-TTS and LibriSpeech sentences. Predicted MOS: 3.85, behind Inworld TTS-2 Flash at 3.88 and ahead of Sonic 3.6 at 3.82. Median time to first byte: 47 ms. Flash v2.5 took 119 ms, Sonic 3.6 143 ms, Inworld TTS-2 Flash 244 ms. Other: 85.5 TTSDS prosody score and 0.64 speaker similarity for cloned audio. Independent results are more mixed. On Cekura’s benchmark of 27 models, TTS-1 scored 81.5% reading accuracy and 4.50 naturalness. Its median streaming latency, timed from the last word sent, was 159 ms. Inworld TTS-2 Flash posted 45 ms and 93.2% accuracy there, at the same $15 price. AssemblyAI also cites a 1.7% word error rate on Coval’s leaderboard. It reports a top-10 debut in Artificial Analysis’ controlled-voice arena. <iframe id="mtp-aai-tts1-frame" title="AssemblyAI-TTS-1 interactive explainer" loading="lazy" style="width:100%;height:600px;border:0;display:block;background:transparent;" scrolling="no" srcdoc="<!doctype html><title>AssemblyAI-TTS-1 Explainer</title>html,body{margin:0;padding:0;background:transparent} #mtp-aai-tts1{--cob:#3923c7;--cob3:#887bdd;--cob2:#b0a7e9;--ink:#1d1b16;--ink2:#2a2721;--line:#4a4945;--cream:#fdfcf8;--muted:#a5a4a2;--blush:#e39389;--mtp:#76B900; max-width:880px!important;margin:0 auto!important;background:var(--ink)!important;color:var(--cream)!important;border:1px solid #3a3730!important;border-radius:16px!important;font-family:Inter,system-ui,-apple-system,Segoe UI,Roboto,sans-serif!important;overflow:hidden;box-sizing:border-box;line-height:1.5} #mtp-aai-tts1 *{box-sizing:border-box} #mtp-aai-tts1 hr,#mtp-aai-tts1 p:empty,#mtp-aai-tts1 del,#mtp-aai-tts1 s{display:none!important} #mtp-aai-tts1 .hd{padding:22px 26px 14px;background:radial-gradient(ellipse at 90% -20%,rgba(57,35,199,.55),transparent 60%)!important} #mtp-aai-tts1 .eyebrow{font:600 11px/1 ui-monospace,Menlo,monospace;letter-spacing:.14em;text-transform:uppercase;color:var(--cob2)!important} #mtp-aai-tts1 h3{margin:8px 0 4px!important;font-size:24px!important;font-weight:700!important;color:var(--cream)!important;border:0!important;padding:0!important} #mtp-aai-tts1 .sub{color:var(--muted)!important;font-size:14px;margin:0!important} #mtp-aai-tts1 .tabs{display:flex;gap:6px;padding:0 26px;border-bottom:1px solid #3a3730!important;overflow-x:auto} #mtp-aai-tts1 .tab{background:transparent!important;border:0!important;border-bottom:2px solid transparent!important;color:var(--muted)!important;font:600 13px Inter,system-ui,sans-serif;padding:12px 10px;cursor:pointer;white-space:nowrap;transition:color .2s} #mtp-aai-tts1 .tab.on{color:var(--cream)!important;border-bottom-color:var(--cob3)!important} #mtp-aai-tts1 .pane{display:none;padding:22px 26px 10px;animation:mtpfade .35s ease} #mtp-aai-tts1 .pane.on{display:block} @keyframes mtpfade{from{opacity:0;transform:translateY(6px)}to{opacity:1;transform:none}} #mtp-aai-tts1 .flow{display:flex;align-items:stretch;gap:8px;margin:4px 0 14px} #mtp-aai-tts1 .node{flex:1;background:var(--ink2)!important;border:1px solid #3a3730!important;border-radius:12px;padding:12px 10px;text-align:center;font-size:12.5px;color:var(--muted)!important;transition:all .35s;position:relative} #mtp-aai-tts1 .node b{display:block;color:var(--cream)!important;font-size:13.5px;margin-bottom:3px} #mtp-aai-tts1 .node.act{border-color:var(--cob3)!important;background:rgba(57,35,199,.28)!important;color:var(--cob2)!important;transform:translateY(-3px);box-shadow:0 8px 24px rgba(57,35,199,.35)} #mtp-aai-tts1 .arrow{align-self:center;color:#4a4945!important;font-size:18px} #mtp-aai-tts1 .wave{display:flex;align-items:center;gap:3px;height:34px;justify-content:center;margin-top:6px} #mtp-aai-tts1 .wave i{display:block;width:4px;border-radius:2px;background:var(--cob3)!important;height:6px;opacity:.35} #mtp-aai-tts1 .playing .wave i{animation:mtpwave 1s ease-in-out infinite;opacity:1} @keyframes mtpwave{0%,100%{height:6px}50%{height:30px}} #mtp-aai-tts1 .detail{background:#24211b!important;border-left:3px solid var(--cob3)!important;border-radius:8px;padding:12px 14px;font-size:14px;min-height:66px;color:var(--cream)!important} #mtp-aai-tts1 .ctrl{display:flex;gap:8px;margin:12px 0 4px;flex-wrap:wrap} #mtp-aai-tts1 .btn{background:var(--cob)!important;color:#fff!important;border:0!important;border-radius:999px;padding:8px 16px;font:600 13px Inter,system-ui,sans-serif;cursor:pointer;transition:background .2s} #mtp-aai-tts1 .btn:hover{background:#614fd2!important} #mtp-aai-tts1 .btn.ghost{background:transparent!important;border:1px solid #4a4945!important;color:var(--cream)!important} #mtp-aai-tts1 .pills{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:14px} #mtp-aai-tts1 .pill{background:transparent!important;border:1px solid #4a4945!important;color:var(--muted)!important;border-radius:999px;padding:6px 12px;font:600 12px Inter,system-ui,sans-serif;cursor:pointer} #mtp-aai-tts1 .pill.on{background:var(--cob)!important;border-color:var(--cob)!important;color:#fff!important} #mtp-aai-tts1 .bars{display:flex;flex-direction:column;gap:9px} #mtp-aai-tts1 .row{display:grid;grid-template-columns:150px 1fr 70px;gap:10px;align-items:center;font-size:13px} #mtp-aai-tts1 .row .nm{color:var(--cream)!important} #mtp-aai-tts1 .track{display:block;height:14px;background:#2a2721!important;border-radius:7px;overflow:hidden} #mtp-aai-tts1 .fill{display:block;height:100%;width:0;border-radius:7px;background:#5a574f!important;transition:width .9s cubic-bezier(.2,.8,.2,1)} #mtp-aai-tts1 .row.me .fill{background:linear-gradient(90deg,#3923c7,#887bdd)!important} #mtp-aai-tts1 .row.me .nm{color:var(--cob2)!important;font-weight:700} #mtp-aai-tts1 .val{text-align:right;font-variant-numeric:tabular-nums;color:var(--cream)!important} #mtp-aai-tts1 .note{font-size:12px;color:var(--muted)!important;margin:12px 0 0!important} #mtp-aai-tts1 .calc label{font-size:13px;color:var(--muted)!important} #mtp-aai-tts1 input[type=range]{width:100%;accent-color:#887bdd;margin:8px 0 4px} #mtp-aai-tts1 .big{font-size:26px;font-weight:700;color:var(--cream)!important} #mtp-aai-tts1 .grid{display:grid;grid-template-columns:1fr 1fr;gap:10px;margin-top:14px;align-items:start} #mtp-aai-tts1 .card{background:var(--ink2)!important;border:1px solid #3a3730!important;border-radius:12px;padding:12px 14px;cursor:pointer;transition:border-color .2s} #mtp-aai-tts1 .card:hover{border-color:var(--cob3)!important} #mtp-aai-tts1 .card h4{margin:0!important;font-size:14px!important;color:var(--cream)!important;display:flex;justify-content:space-between} #mtp-aai-tts1 .card p{margin:8px 0 0!important;font-size:13px;color:var(--muted)!important;display:none} #mtp-aai-tts1 .card.open p{display:block;animation:mtpfade .3s ease} #mtp-aai-tts1 .card .tag{font:600 10px ui-monospace,monospace;color:var(--blush)!important;letter-spacing:.08em} #mtp-aai-tts1 .cost{display:grid;grid-template-columns:1fr auto;gap:6px 12px;font-size:14px;margin-top:10px} #mtp-aai-tts1 .cost .me{color:var(--cob2)!important;font-weight:700} #mtp-aai-tts1 .ft{display:flex;justify-content:space-between;gap:10px;flex-wrap:wrap;padding:14px 26px 18px;font-size:11.5px;color:var(--muted)!important;border-top:1px solid #3a3730!important;margin-top:14px} #mtp-aai-tts1 .ft .mtp{color:var(--mtp)!important;font-weight:700} #mtp-aai-tts1 a{color:var(--cob2)!important} @media (max-width:640px){ #mtp-aai-tts1 .hd,#mtp-aai-tts1 .pane{padding-left:16px;padding-right:16px} #mtp-aai-tts1 .tabs{padding:0 12px} #mtp-aai-tts1 .flow{flex-direction:column} #mtp-aai-tts1 .arrow{transform:rotate(90deg)} #mtp-aai-tts1 .row{grid-template-columns:110px 1fr 58px;font-size:12px} #mtp-aai-tts1 .grid{grid-template-columns:1fr} #mtp-aai-tts1 h3{font-size:20px!important} } <div id="mtp-aai-tts1"> <div class="hd"> <div class="eyebrow">AssemblyAI · Text-to-Speech · Oct 2026</div> <h3>AssemblyAI-TTS-1, explained</h3> <p class="sub">A streaming voice for phone agents: $15 per 1M characters, 4,000+ voices, 6 languages.</p> </div> <div class="tabs" role="tablist"> <button class="tab on" data-t="flow">How it works</button> <button class="tab" data-t="bench">Benchmarks</button> <button class="tab" data-t="cost">Cost calculator</button> <button class="tab" data-t="limits">Limits</button> </div> <div class="pane on" data-p="flow"> <div class="flow" id="aai-flow"> <div class="node" data-i="0"><b>Agent LLM</b>writes the reply</div><div class="arrow">›</div> <div class="node" data-i="1"><b>WebSocket</b>1 session per call</div><div class="arrow">›</div> <div class="node" data-i="2"><b>TTS-1</b>continuous audio</div><div class="arrow">›</div> <div class="node" data-i="3"><b>Audio out</b>8 to 48 kHz<div class="wave"><i></i><i></i><i></i><i></i><i></i><i></i><i></i></div></div><div class="arrow">›</div> <div class="node" data-i="4"><b>Barge-in</b>cancel message</div> </div> <div class="detail" id="aai-detail">Press Play or Next to step through a live call turn.</div> <div class="ctrl"> <button class="btn" id="aai-play"> Play</button> <button class="btn ghost" id="aai-prev">‹ Back</button> <button class="btn ghost" id="aai-next">Next ›</button> </div> </div> <div class="pane" data-p="bench"> <div class="pills" id="aai-pills"> <button class="pill on" data-m="ent">Entity accuracy</button> <button class="pill" data-m="ttfb">Median TTFB (AssemblyAI test)</button> <button class="pill" data-m="cstream">Streaming latency (Cekura)</button> <button class="pill" data-m="cacc">Reading accuracy (Cekura)</button> </div> <div class="bars" id="aai-bars"></div> <p class="note" id="aai-bnote"></p> </div> <div class="pane" data-p="cost"> <div class="calc"> <label for="aai-min">Spoken minutes per month: <b id="aai-minv" style="color:#fdfcf8">10,000</b></label> <div class="cost" id="aai-cost"></div> <p class="note">Assumes 600 characters per minute of phone speech, the figure AssemblyAI cites as typical. List prices per 1M characters: AssemblyAI-TTS-1 $15, Inworld TTS-2 Flash $15 on demand, ElevenLabs Flash v2.5 $40, ElevenLabs v4 $80. Concurrency limits are not priced in.</p> </div> </div> <div class="pane" data-p="limits"> <div class="grid" id="aai-cards"> <div class="card"><h4>No SSML or speed control <span class="tag">LAUNCH GAP</span></h4><p>The WebSocket API has no speed, pitch, emotion or style parameter, and neither API accepts SSML. Punctuation and written fillers like "um" still shape delivery.</p></div> <div class="card"><h4>No pronunciation dictionaries <span class="tag">COMING SOON</span></h4><p>Custom pronunciation dictionaries and emotion tags are announced but not shipped. Cartesia Sonic 3.6 already offers custom pronunciations.</p></div> <div class="card"><h4>6 languages <span class="tag">COVERAGE</span></h4><p>English, Spanish, French, German, Italian and Portuguese. ElevenLabs Flash v2.5 lists 32 and Cartesia Sonic 3.6 lists 44.</p></div> <div class="card"><h4>Latency depends on method <span class="tag">CHECK IT</span></h4><p>About 90 ms advertised, 47 ms median in AssemblyAI's own test, 159 ms median streaming latency on Cekura (timed from the last word sent).</p></div> <div class="card"><h4>Instant cloning only <span class="tag">LAUNCH GAP</span></h4><p>Pro (fine-tuned) voice cloning is not available yet. Instant clones need a short clip and speaker permission.</p></div> <div class="card"><h4>Cloud only <span class="tag">DEPLOYMENT</span></h4><p>US (us-east-1) and EU (eu-north-1) hosts at the same price. No on-prem option is announced.</p></div> </div> <p class="note">Tap a card to expand it.</p> </div> <div class="ft"><span>Sources: AssemblyAI launch post, docs and pricing; Cekura TTS benchmark; vendor pricing pages. Checked Oct 29, 2026.</span><span class="mtp">Built by Marktechpost</span></div> </div> (function(){ var R=document.getElementById('mtp-aai-tts1');if(!R)return; function resize(){try{parent.postMessage({mtpEmbed:'aai-tts1',h:R.offsetHeight+40},'*')}catch(e){}} R.querySelectorAll('.tab').forEach(function(t){t.addEventListener('click',function(){ R.querySelectorAll('.tab').forEach(function(x){x.classList.remove('on')});t.classList.add('on'); R.querySelectorAll('.pane').forEach(function(p){p.classList.toggle('on',p.getAttribute('data-p')===t.getAttribute('data-t'))}); if(t.getAttribute('data-t')==='bench')draw(cur);setTimeout(resize,60);});}); var steps=['Your agent LLM writes a reply. Text can be sent word by word as it is generated.', 'One WebSocket session per conversation. Voice, language, sample rate and encoding are fixed at connect time.', 'TTS-1 models speech on continuous audio representations, not discrete tokens. AssemblyAI says this improves realism and speed.', 'Audio streams back as raw mono PCM, mu-law or A-law. 24 kHz PCM is native; 8 kHz mu-law suits phone lines.', 'If the caller interrupts, a cancel message stops speech. Word timestamps show what was actually said.']; var nodes=R.querySelectorAll('#aai-flow .node'),det=R.querySelector('#aai-detail'),si=-1,timer=null; function show(i){si=(i+nodes.length)%nodes.length;nodes.forEach(function(n,k){n.classList.toggle('act',k===si);n.classList.toggle('playing',k===3&&si>=3)});det.textContent=(si+1)+'/5 · '+steps[si];resize();} R.querySelector('#aai-next').onclick=function(){stop();show(si+1)}; R.querySelector('#aai-prev').onclick=function(){stop();show(si<=0?4:si-1)}; var pb=R.querySelector('#aai-play'); function stop(){if(timer){clearInterval(timer);timer=null;pb.textContent=' Play'}} pb.onclick=function(){if(timer){stop();return}pb.textContent='❚❚ Pause';show(si+1);timer=setInterval(function(){show(si+1)},2200)}; var D={ ent:{u:'%',max:100,hi:true,note:'AssemblyAI pronunciation set: 580 hard targets (phone numbers, dates, URLs, codes). Self-reported, eval dated Oct 28, 2026.', r:[['AssemblyAI-TTS-1',93.7,1],['ElevenLabs v4',93.6],['Cartesia Sonic 3.6',82.5],['ElevenLabs Flash v2.5',68.0]]}, ttfb:{u:' ms',max:650,hi:false,note:'Median time to first byte over each API from one client, one request at a time. Self-reported by AssemblyAI. Lower is better.', r:[['AssemblyAI-TTS-1',47,1],['ElevenLabs Flash v2.5',119],['Cartesia Sonic 3.6',143],['Inworld TTS-2 Flash',244],['Inworld TTS-2',280],['ElevenLabs v4',612]]}, cstream:{u:' ms',max:600,hi:false,note:'Cekura: median from the last word sent to first audible sound, US East, 104 sentences x 3 reads. Independent. Lower is better.', r:[['Inworld TTS-2 Flash',45],['AssemblyAI-TTS-1',159,1],['ElevenLabs Flash v2.5',177],['Cartesia Sonic 3.6',270],['ElevenLabs v4',561]]}, cacc:{u:'%',max:100,hi:true,note:'Cekura: hard parts read correctly, scored by speech-to-text transcribers (an estimate, not human-rated). Independent.', r:[['Inworld TTS-2 Flash',93.2],['AssemblyAI-TTS-1',81.5,1],['ElevenLabs v4',80.4],['Cartesia Sonic 3.6',74.9],['ElevenLabs Flash v2.5',64.4]]}}; var cur='ent',bars=R.querySelector('#aai-bars'),bn=R.querySelector('#aai-bnote'); function draw(m){cur=m;var d=D[m];bars.innerHTML='';d.r.forEach(function(x){ var row=document.createElement('div');row.className='row'+(x[2]?' me':''); row.innerHTML='<span class="nm">'+x[0]+'</span><span class="track"><span class="fill"></span></span><span class="val">'+(d.u==='%'?x[1].toFixed(1):x[1])+d.u+'</span>'; bars.appendChild(row);var f=row.querySelector('.fill');setTimeout(function(){f.style.width=Math.max(2,x[1]/d.max*100)+'%'},40);}); bn.textContent=d.note;setTimeout(resize,60);} R.querySelectorAll('#aai-pills .pill').forEach(function(p){p.addEventListener('click',function(){ R.querySelectorAll('#aai-pills .pill').forEach(function(x){x.classList.remove('on')});p.classList.add('on');draw(p.getAttribute('data-m'))})}); var rng=R.querySelector('#aai-min'),mv=R.querySelector('#aai-minv'),co=R.querySelector('#aai-cost'); var P=[['AssemblyAI-TTS-1',15,1],['Inworld TTS-2 Flash (on demand)',15],['ElevenLabs Flash v2.5',40],['ElevenLabs v4',80]]; function calc(){var m=+rng.value,ch=m*600;mv.textContent=m.toLocaleString('en-US');var h=''; P.forEach(function(p){var c=ch/1e6*p[1];h+='<span class="&apos;+(p[2]?&apos;me&apos;:&apos;&apos;)+&apos;">'+p[0]+'</span><span class="&apos;+(p[2]?&apos;me&apos;:&apos;&apos;)+&apos;">$'+c.toLocaleString('en-US',{maximumFractionDigits:0})+'/mo</span>'}); co.innerHTML=h+'<span style="color:#a5a4a2">Characters per month</span><span style="color:#a5a4a2">'+(ch/1e6).toLocaleString('en-US',{maximumFractionDigits:1})+'M</span>';} rng.addEventListener('input',calc);calc(); R.querySelectorAll('#aai-cards .card').forEach(function(c){c.addEventListener('click',function(){c.classList.toggle('open');setTimeout(resize,40)})}); draw('ent');show(0); window.addEventListener('load',resize);window.addEventListener('resize',resize);setTimeout(resize,300); })(); "> (function(){var f=document.getElementById("mtp-aai-tts1-frame");window.addEventListener("message",function(e){if(e.data&&e.data.mtpEmbed==="aai-tts1"&&e.data.h){f.style.height=e.data.h+"px";}});})(); How does AssemblyAI-TTS-1 compare with ElevenLabs, Cartesia and Inworld? Cartesia’s range divides each plan price by its credits, at 1 credit per character. What does AssemblyAI-TTS-1 cost? Pricing is per input character, including spaces and punctuation, at $15 per 1M. Every voice, cloning, word timestamps and EU processing come at that same rate. There are no per-voice fees, concurrency fees or minimums. AssemblyAI calls it the lowest-priced frontier TTS model for voice agents. However, Inworld TTS-2 Flash also lists $15 per 1M on demand, falling to $7 on paid plans. The sharper difference is concurrency: Inworld’s on-demand tier allows 5 concurrent requests. Key Takeaways AssemblyAI-TTS-1 is AssemblyAI’s first text-to-speech model, built for live voice agents. It costs $15 per 1M characters, with unlimited concurrency and no contract. Self-reported: 93.7% entity accuracy and 47 ms median time to first byte. Cekura measured 159 ms streaming latency and 81.5% reading accuracy. Missing at launch: SSML, speed control, pronunciation dictionaries and emotion tags. Check out the launch post, docs, pricing, and the LiveKit and Pipecat reference agents. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post AssemblyAI Ships AssemblyAI-TTS-1: A $15 per 1M Character Text-to-Speech API for Voice Agents appeared first on MarkTechPost.

AI

NVIDIA Releases Nemotron 3.5 Transcribe: An Open 30B MoE Model for 40-Language Transcription with ~4B Active Parameters

NVIDIA has released Nemotron 3.5 Transcribe, an open-weights speech LLM built for transcription. One model handles 40 languages and recordings up to 1 hour. A text instruction picks the output style for each request. Options include verbatim, written form, speaker-attributed, contextual and translated output. For developers, it can replace separate ASR, diarization and alignment tools. TL;DR Size: 34B total and about 4.3B active parameters; 65,536-token context; audio up to 1 hour. Runs on: 1x H100 80 GB in BF16. FP8 (about 35 GB) and NVFP4 (about 21 GB) checkpoints exist; NVFP4 runs on an RTX 5090. GGUF files target NeMo-Speech.cpp. Performance: Best open-weights model on the English and multilingual Open ASR Leaderboards; top commercial APIs still lead. Best: 4.71% mean WER on the 9 public English Open ASR sets (next open model: 4.94%). Worst: 34.95% cpCER on AliMeeting Mandarin far-field meetings. Bottom line: Best thing: 1 open model covers transcripts, speaker labels, timestamps and translation. Worst thing: offline only, with weak Mandarin meeting accuracy. What is Nemotron 3.5 Transcribe? Nemotron 3.5 Transcribe is a speech-augmented language model (SALM) from NVIDIA that turns audio and instructions into text. An audio encoder feeds embeddings into an LLM decoder, which makes transcription promptable. NVIDIA’s technical research report argues there is no single correct transcript. Numbers, fillers and domain terms depend on who reads the output. NVIDIA positions the model as a transcription specialist, not a general-purpose assistant. How does the architecture work? It has three parts: Parallel Expert Encoder: a trainable 0.6B Transformer ASR encoder runs beside a frozen Nemotron 3 Diarization model. The diarizer tracks up to 8 speakers. Its speaker activity is added to the ASR features every 80 ms. The LLM therefore sees who spoke when, without architecture changes. LLM backbone: Nemotron 3.5 Lightning 30B-A3B, a hybrid Mamba-Transformer MoE. It has 52 layers: 23 Mamba-2, 23 MoE and only 6 attention layers. Each MoE layer routes to 6 of 128 experts plus 1 shared expert. Few attention layers keep the KV cache small for hour-long audio. Add-ons: a shared multi-token prediction (MTP) layer drafts tokens for speculative decoding. A 50M-parameter CTC adapter aligns the finished transcript to audio for word timestamps. The model name says 30B-A4B, but the checkpoint holds about 34B parameters. The extra weight comes from the 1.3B MTP module and the audio components. Training used 1 supervised fine-tuning stage on 256 H100 GPUs for 28,500 steps. A short RL stage followed: 400 GSPO updates with a reward of 1 minus WER. The model saw about 1.72M unique hours of audio, or 6.5M hours with repeats. How does it score on benchmarks? All numbers below are self-reported by NVIDIA in the technical report. English Open ASR (9 public sets): 4.71% mean WER. The next open model, Qwen3-ASR-1.7B, scores 4.94%. Seven commercial systems still lead, topped by OtterAI speech-v3.1 at 3.88%. Multilingual Open ASR (7 languages): 4.62% mean WER, versus 6.49% for Qwen3-ASR-1.7B. ElevenLabs Scribe v2 reaches 3.56%. Long-form ASR: 10.98% mean WER, versus 13.37% for Parakeet-TDT-0.6B-v3. Multi-speaker, full sessions: 15.80% mean cpWER, versus 16.83% for DiCoW v3.3 with DiariZen. NVIDIA says it is the first SALM to beat cascaded systems here. Contextual ASR: an entity list cuts named-entity WER on ContextASR-Bench from 12.21% to 2.20%. <iframe id="mtp-nt35-frame" title="Nemotron 3.5 Transcribe explainer" style="width:100%;max-width:880px;height:600px;border:0;display:block;margin:0 auto;" loading="lazy" srcdoc="<!doctype html><title>Nemotron 3.5 Transcribe Explainer</title>html,body{margin:0;background:#0d0d0d}<div id="mtp-nt35"> #mtp-nt35{--g:#76B900;--g2:#5a8f00;--bg:#0d0d0d;--card:#161616;--line:#2a2a2a;--tx:#ededed;--mu:#9a9a9a;font-family:Inter,-apple-system,Segoe UI,Roboto,sans-serif!important;background:var(--bg)!important;color:var(--tx)!important;max-width:880px;margin:0 auto;border:1px solid var(--line)!important;border-radius:14px;overflow:hidden;box-sizing:border-box} #mtp-nt35 *{box-sizing:border-box} #mtp-nt35 hr,#mtp-nt35 p:empty,#mtp-nt35 del,#mtp-nt35 s{display:none!important} #mtp-nt35 .hd{padding:22px 24px 14px;border-bottom:1px solid var(--line)!important;background:linear-gradient(135deg,#121a05 0%,#0d0d0d 60%)!important} #mtp-nt35 .tag{display:inline-block;font-size:11px;letter-spacing:.12em;text-transform:uppercase;color:var(--g)!important;border:1px solid var(--g)!important;border-radius:999px;padding:3px 10px;margin-bottom:10px} #mtp-nt35 h3{margin:0 0 4px!important;font-size:22px!important;line-height:1.25!important;color:#fff!important;font-weight:800!important} #mtp-nt35 .sub{margin:0!important;color:var(--mu)!important;font-size:14px!important} #mtp-nt35 .tabs{display:flex;gap:4px;padding:10px 12px 0;border-bottom:1px solid var(--line)!important;overflow-x:auto} #mtp-nt35 .tab{background:transparent!important;border:0!important;color:var(--mu)!important;font:600 13px Inter,sans-serif;padding:10px 14px;cursor:pointer;border-bottom:2px solid transparent!important;white-space:nowrap;transition:color .2s} #mtp-nt35 .tab:hover{color:#fff!important} #mtp-nt35 .tab.on{color:var(--g)!important;border-bottom-color:var(--g)!important} #mtp-nt35 .pane{display:none;padding:20px 24px 22px;animation:nt35fade .35s ease} #mtp-nt35 .pane.on{display:block} @keyframes nt35fade{from{opacity:0;transform:translateY(6px)}to{opacity:1;transform:none}} #mtp-nt35 .flow{display:grid;grid-template-columns:repeat(4,1fr);gap:10px;position:relative} #mtp-nt35 .node{background:var(--card)!important;border:1px solid var(--line)!important;border-radius:10px;padding:12px;min-height:96px;transition:all .35s;cursor:pointer;position:relative} #mtp-nt35 .node b{display:block;font-size:13px;color:#fff!important;margin-bottom:4px} #mtp-nt35 .node span{font-size:12px;color:var(--mu)!important;line-height:1.4} #mtp-nt35 .node .n{position:absolute;top:8px;right:10px;font-size:11px;color:var(--mu)!important} #mtp-nt35 .node.on{border-color:var(--g)!important;box-shadow:0 0 0 1px var(--g),0 0 22px rgba(118,185,0,.25);transform:translateY(-2px)} #mtp-nt35 .node.on .n{color:var(--g)!important} #mtp-nt35 .wire{height:4px!important;background:var(--line)!important;border-radius:4px;margin:14px 0 4px;overflow:hidden} #mtp-nt35 .wire i{display:block;height:4px!important;width:0;background:var(--g)!important;transition:width .5s ease} #mtp-nt35 .detail{background:#111!important;border:1px solid var(--line)!important;border-left:3px solid var(--g)!important;border-radius:8px;padding:14px 16px;margin-top:12px;font-size:14px;line-height:1.55;min-height:92px} #mtp-nt35 .detail strong{color:var(--g)!important} #mtp-nt35 .ctrl{display:flex;gap:8px;margin-top:12px;flex-wrap:wrap} #mtp-nt35 .btn{background:#1d1d1d!important;color:#fff!important;border:1px solid var(--line)!important;border-radius:8px;padding:8px 14px;font:600 13px Inter,sans-serif;cursor:pointer;transition:all .2s} #mtp-nt35 .btn:hover{border-color:var(--g)!important} #mtp-nt35 .btn.pri{background:var(--g)!important;color:#0b0b0b!important;border-color:var(--g)!important} #mtp-nt35 .seg{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:14px} #mtp-nt35 .chip{background:#1a1a1a!important;color:var(--mu)!important;border:1px solid var(--line)!important;border-radius:999px;padding:6px 12px;font:600 12px Inter,sans-serif;cursor:pointer;transition:all .2s} #mtp-nt35 .chip.on{color:#0b0b0b!important;background:var(--g)!important;border-color:var(--g)!important} #mtp-nt35 .bar{display:grid;grid-template-columns:170px 1fr 58px;align-items:center;gap:10px;margin:9px 0;font-size:13px} #mtp-nt35 .bar .lb{color:#ddd!important;white-space:nowrap;overflow:hidden;text-overflow:ellipsis} #mtp-nt35 .bar .tr{background:#1c1c1c!important;border-radius:6px;height:22px!important;overflow:hidden} #mtp-nt35 .bar .fl{height:22px!important;width:0;background:#3a3a3a!important;border-radius:6px;transition:width .9s cubic-bezier(.2,.8,.2,1)} #mtp-nt35 .bar.me .fl{background:linear-gradient(90deg,var(--g2),var(--g))!important} #mtp-nt35 .bar.me .lb{color:var(--g)!important;font-weight:700} #mtp-nt35 .bar .v{text-align:right;font-variant-numeric:tabular-nums;color:#fff!important;font-weight:700} #mtp-nt35 .note{font-size:12px;color:var(--mu)!important;margin-top:10px!important;line-height:1.5} #mtp-nt35 .grid2{display:grid;grid-template-columns:1fr 1fr;gap:10px} #mtp-nt35 .card{background:var(--card)!important;border:1px solid var(--line)!important;border-radius:10px;padding:14px;transition:all .25s;cursor:pointer} #mtp-nt35 .card:hover,#mtp-nt35 .card.on{border-color:var(--g)!important} #mtp-nt35 .card h4{margin:0 0 6px!important;font-size:15px!important;color:#fff!important} #mtp-nt35 .card .big{font-size:24px;font-weight:800;color:var(--g)!important} #mtp-nt35 .card p{margin:4px 0 0!important;font-size:13px!important;color:var(--mu)!important;line-height:1.45} #mtp-nt35 .card .more{max-height:0;overflow:hidden;transition:max-height .35s ease;font-size:13px;color:#ccc!important;line-height:1.45} #mtp-nt35 .card.on .more{max-height:120px;margin-top:8px} #mtp-nt35 .out{background:#0a0a0a!important;border:1px solid var(--line)!important;border-radius:8px;padding:14px;margin-top:10px} #mtp-nt35 pre,#mtp-nt35 code{background:#0a0a0a!important;color:#d8f5a8!important;border:0!important;font:13px/1.55 ui-monospace,SFMono-Regular,Menlo,monospace!important;white-space:pre-wrap!important;margin:0!important;padding:0!important} #mtp-nt35 .lbl{font-size:11px;letter-spacing:.1em;text-transform:uppercase;color:var(--mu)!important;margin-bottom:6px} #mtp-nt35 .row{display:flex;gap:10px;align-items:flex-start;padding:10px 0;border-bottom:1px solid #1f1f1f!important;font-size:14px;line-height:1.5} #mtp-nt35 .row:last-child{border-bottom:0!important} #mtp-nt35 .ic{flex:0 0 22px;height:22px;border-radius:6px;display:flex;align-items:center;justify-content:center;font-weight:800;font-size:13px} #mtp-nt35 .ic.up{background:rgba(118,185,0,.15)!important;color:var(--g)!important} #mtp-nt35 .ic.dn{background:rgba(255,120,80,.15)!important;color:#ff8a5c!important} #mtp-nt35 .ft{display:flex;justify-content:space-between;flex-wrap:wrap;gap:6px;padding:12px 24px;border-top:1px solid var(--line)!important;font-size:12px;color:var(--mu)!important} #mtp-nt35 .ft a,#mtp-nt35 .ft .br{color:var(--g)!important;text-decoration:none;font-weight:700} @media(max-width:640px){ #mtp-nt35 .hd{padding:18px 16px 12px} #mtp-nt35 h3{font-size:18px!important} #mtp-nt35 .pane{padding:16px} #mtp-nt35 .flow{grid-template-columns:1fr 1fr} #mtp-nt35 .grid2{grid-template-columns:1fr} #mtp-nt35 .bar{grid-template-columns:110px 1fr 50px;font-size:12px} #mtp-nt35 .ft{padding:12px 16px} } <div class="hd"> <div class="tag">NVIDIA · Open weights · Oct 28, 2026</div> <h3>How Nemotron 3.5 Transcribe works</h3> <p class="sub">34B total, ~4.3B active parameters · 40 languages · audio up to 1 hour</p> </div> <div class="tabs" role="tablist"> <button class="tab on" data-t="p1">How it works</button> <button class="tab" data-t="p2">Benchmarks</button> <button class="tab" data-t="p3">Prompt styles</button> <button class="tab" data-t="p4">Deploy</button> <button class="tab" data-t="p5">Trade-offs</button> </div> <div class="pane on" id="p1"> <div class="flow"> <div class="node on" data-s="0"><span class="n">01</span><b>Audio in</b><span>16 kHz mono, split into 30 s encoder chunks</span></div> <div class="node" data-s="1"><span class="n">02</span><b>Parallel Expert Encoder</b><span>0.6B ASR encoder + frozen diarizer</span></div> <div class="node" data-s="2"><span class="n">03</span><b>Hybrid MoE LLM</b><span>Nemotron 3.5 Lightning 30B-A3B</span></div> <div class="node" data-s="3"><span class="n">04</span><b>Text + timestamps</b><span>CTC adapter aligns words to audio</span></div> </div> <div class="wire"><i id="nt35w"></i></div> <div class="detail" id="nt35d"></div> <div class="ctrl"> <button class="btn" id="nt35prev">← Back</button> <button class="btn pri" id="nt35next">Next step →</button> <button class="btn" id="nt35play"> Auto-play</button> </div> </div> <div class="pane" id="p2"> <div class="seg"> <button class="chip on" data-b="en">English Open ASR</button> <button class="chip" data-b="ml">Multilingual Open ASR</button> <button class="chip" data-b="lf">Long-form</button> </div> <div id="nt35bars"></div> <p class="note" id="nt35bn"></p> </div> <div class="pane" id="p3"> <p class="note" style="margin:0 0 10px!important">1 checkpoint, different instructions. Click a style to see the request pattern and output shape (illustrative, not a model sample).</p> <div class="seg" id="nt35ps"> <button class="chip on" data-p="0">Verbatim</button> <button class="chip" data-p="1">Written form</button> <button class="chip" data-p="2">Context terms</button> <button class="chip" data-p="3">Speakers</button> <button class="chip" data-p="4">Translate</button> </div> <div class="out"><div class="lbl">Instruction</div><pre><code id="nt35pi"></code></pre></div> <div class="out"><div class="lbl">Output shape</div><pre><code id="nt35po"></code></pre></div> </div> <div class="pane" id="p4"> <div class="grid2"> <div class="card on"><h4>BF16</h4><div class="big">67.5 GB</div><p>Reference weights · 1x H100 80 GB</p><div class="more">Ampere, Hopper and Blackwell. NVIDIA estimates about 410 RTFx on 1 H100 with MTP2 (batched English ASR).</div></div> <div class="card"><h4>FP8</h4><div class="big">~35 GB</div><p>Ada Lovelace, Hopper, Blackwell</p><div class="more">Error rate 1.0% higher in relative terms on average. NVIDIA recommends it when non-English accuracy matters most.</div></div> <div class="card"><h4>NVFP4</h4><div class="big">~21 GB</div><p>Blackwell only · runs on RTX 5090</p><div class="more">Error rate 3.6% higher in relative terms on average. Best when memory or throughput is the main constraint.</div></div> <div class="card"><h4>NIM speed</h4><div class="big">2,238x</div><p>Real time on 1 RTX PRO 6000 Blackwell</p><div class="more">NVFP4 + DFlash2 in the Nemotron Transcribe NIM; 30 s clip in 0.13 s. These serving optimizations are not open source yet.</div></div> </div> <p class="note">Runtimes: vLLM 0.32.0 + NeMo Speech plugin, Hugging Face Transformers, SGLang-Omni 0.2.0, GGML (NeMo-Speech.cpp). License: OpenMDW-1.1.</p> </div> <div class="pane" id="p5"> <div class="row"><div class="ic up">+</div><div><b>Best open-weights ASR:</b> 4.71% English and 4.62% multilingual mean WER on the Open ASR Leaderboards.</div></div> <div class="row"><div class="ic up">+</div><div><b>Meetings:</b> 15.80% mean cpWER on full sessions, ahead of DiCoW v3.3 + DiariZen at 16.83%.</div></div> <div class="row"><div class="ic up">+</div><div><b>Context:</b> an entity list cuts named-entity WER on ContextASR-Bench from 12.21% to 2.20%.</div></div> <div class="row"><div class="ic dn">!</div><div><b>Commercial gap:</b> OtterAI speech-v3.1 reaches 3.88% English WER; ElevenLabs Scribe v2 reaches 3.56% multilingual.</div></div> <div class="row"><div class="ic dn">!</div><div><b>Mandarin meetings:</b> 32.91% cpCER on AISHELL-4 and 34.95% on AliMeeting.</div></div> <div class="row"><div class="ic dn">!</div><div><b>Offline only:</b> not a streaming ASR service; release-day vLLM is roughly 2x slower than the NIM.</div></div> </div> <div class="ft"><span>Source: NVIDIA technical report and model card (self-reported), Oct 2026</span><span class="br">© Marktechpost</span></div> (function(){ var R=document.getElementById('mtp-nt35'); function post(){try{parent.postMessage({mtpEmbed:'nt35',h:R.offsetHeight+40},'*')}catch(e){}} R.querySelectorAll('.tab').forEach(function(t){t.addEventListener('click',function(){ R.querySelectorAll('.tab').forEach(function(x){x.classList.remove('on')}); R.querySelectorAll('.pane').forEach(function(x){x.classList.remove('on')}); t.classList.add('on');document.getElementById(t.dataset.t).classList.add('on'); if(t.dataset.t==='p2')drawBars(curB);setTimeout(post,60);setTimeout(post,500);});}); var S=[ '<strong>Audio in.</strong> The model takes 16 kHz single-channel audio, up to 1 hour per request. Encoders process independent 30 s chunks, so encoder compute grows linearly with length. Each second of audio becomes 12.5 LLM input positions.', '<strong>Parallel Expert Encoder.</strong> A trainable 32-layer Transformer ASR encoder (0.6B) runs beside the frozen Nemotron 3 Diarization model (~100M, up to 8 speakers). Speaker activity is added to ASR features every 80 ms, so the LLM sees who spoke when.', '<strong>Hybrid MoE LLM.</strong> Nemotron 3.5 Lightning 30B-A3B: 23 Mamba-2, 23 MoE and 6 attention layers. 128 experts, top-6 routing, 1 shared expert, ~3.5B active. A shared MTP layer drafts tokens for speculative decoding.', '<strong>Text + timestamps.</strong> The LLM writes the transcript in the style the instruction asks for. A 50M-parameter CTC adapter then force-aligns words to audio, giving speaker-attributed word timestamps.']; var cur=0,timer=null,nodes=R.querySelectorAll('.node'),D=document.getElementById('nt35d'),W=document.getElementById('nt35w'); function go(i){cur=(i+4)%4;nodes.forEach(function(n,k){n.classList.toggle('on',k===cur)});D.innerHTML=S[cur];W.style.width=((cur+1)*25)+'%';} nodes.forEach(function(n){n.addEventListener('click',function(){stop();go(+n.dataset.s)})}); document.getElementById('nt35next').onclick=function(){stop();go(cur+1)}; document.getElementById('nt35prev').onclick=function(){stop();go(cur-1)}; var pb=document.getElementById('nt35play'); function stop(){if(timer){clearInterval(timer);timer=null;pb.textContent=' Auto-play'}} pb.onclick=function(){if(timer){stop();return}pb.textContent='❚❚ Pause';timer=setInterval(function(){go(cur+1)},2200)}; go(0); var B={ en:{d:[['Nemotron 3.5 Transcribe',4.71,1],['Qwen3-ASR-1.7B',4.94],['Canary-Qwen-2.5B',5.38],['Parakeet-TDT-0.6B-v3',5.58],['Whisper-large-v3',6.66]],m:8,n:'Mean WER (%) on the 9 public English Open ASR sets, lower is better. Baselines: leaderboard results of Oct 21, 2026, as listed in NVIDIA's report.'}, ml:{d:[['Nemotron 3.5 Transcribe',4.62,1],['Qwen3-ASR-1.7B',6.49],['Voxtral-Small-24B',6.52],['Whisper-large-v3',7.94]],m:9,n:'Mean WER (%) over 7 languages (de, fr, it, es, pt, nl, hi; Hindi uses OIWER), lower is better. Baselines: leaderboard results of Oct 2, 2026.'}, lf:{d:[['Nemotron 3.5 Transcribe',10.98,1],['Parakeet-TDT-0.6B-v3',13.37],['Whisper-large-v3',13.92],['Canary-Qwen-2.5B',14.08]],m:16,n:'Mean WER (%) on the long-form track (Earnings-21, Earnings-22, CORAAL), lower is better.'}}; var curB='en',BB=document.getElementById('nt35bars'),BN=document.getElementById('nt35bn'); function drawBars(k){curB=k;var b=B[k];BB.innerHTML=b.d.map(function(r){return '<div class="bar&apos;+(r[2]?&apos; me&apos;:&apos;&apos;)+&apos;"><div class="lb">'+r[0]+'</div><div class="tr"><div class="fl" data-w="&apos;+(r[1]/b.m*100)+&apos;"></div></div><div class="v">'+r[1].toFixed(2)+'%</div></div>'}).join('');BN.textContent=b.n; setTimeout(function(){BB.querySelectorAll('.fl').forEach(function(f){f.style.width=f.dataset.w+'%'})},40);setTimeout(post,80);} R.querySelectorAll('[data-b]').forEach(function(c){c.addEventListener('click',function(){R.querySelectorAll('[data-b]').forEach(function(x){x.classList.remove('on')});c.classList.add('on');drawBars(c.dataset.b)})}); drawBars('en'); var P=[ ['Audio transcription. Language: as spoken.nWording: verbatim, including fillers and repetitions.nFormatting: punctuation and capitalization.nOutput: transcript only.','So, uh, the the launch moved tonnext Tuesday at three thirty.'], ['Audio transcription. Language: as spoken.nWording: omit fillers and accidental repetitions.nNumbers in their standard written form.','So, the launch moved tonnext Tuesday at 3:30.'], ['Transcribe with punctuation and capitalization.nContext terms: Nemotron, NVFP4, DFlash2.','We tested Nemotron in NVFP4nwith the DFlash2 draft model.'], ['Transcribe this audio with punctuation and capitalization.nPut speaker tags non the transcription.n(+ timestamps enabled in the request)',' Can we ship on Friday?n Yes, if QA signs off.nnwords: [{"word":"Can","start":0.42,"end":0.58,"speaker":"spk:0"}, ...]'], ['Please translate the given speech to German.','Der Start wurde auf nächstennDienstag um 15:30 Uhr verschoben.']]; var PI=document.getElementById('nt35pi'),PO=document.getElementById('nt35po'),tt=null; function showP(i){PI.textContent=P[i][0];var s=P[i][1],k=0;PO.textContent='';if(tt)clearInterval(tt);tt=setInterval(function(){k+=2;PO.textContent=s.slice(0,k);if(k>=s.length){clearInterval(tt);post()}},16);setTimeout(post,60);} R.querySelectorAll('[data-p]').forEach(function(c){c.addEventListener('click',function(){R.querySelectorAll('[data-p]').forEach(function(x){x.classList.remove('on')});c.classList.add('on');showP(+c.dataset.p)})}); showP(0); R.querySelectorAll('.card').forEach(function(c){c.addEventListener('click',function(){R.querySelectorAll('.card').forEach(function(x){if(x!==c)x.classList.remove('on')});c.classList.toggle('on');setTimeout(post,380)})}); window.addEventListener('load',post);window.addEventListener('resize',post);setTimeout(post,300); })(); </div> "> (function(){var f=document.getElementById("mtp-nt35-frame");window.addEventListener("message",function(e){if(e.data&&e.data.mtpEmbed==="nt35"&&e.data.h){f.style.height=e.data.h+"px";}});})(); <!-- Fallback: upload nemotron-transcribe-explainer-standalone.html to /wp-content/uploads/ via FTP and use: --> How do you deploy Nemotron 3.5 Transcribe? Weights ship in BF16, FP8, NVFP4 and GGUF. A DFlash2 draft model is optional. The OpenMDW-1.1 license allows commercial use. Runtimes include vLLM 0.32.0 with the NeMo Speech plugin, Transformers and SGLang-Omni 0.2.0. vLLM exposes an OpenAI-compatible API. NVIDIA model’s main speed comes from its Nemotron Transcribe NIM. With NVFP4 and DFlash2, 1 RTX PRO 6000 Blackwell Server Edition transcribes 2,238 seconds of audio per second. It returns a 30-second clip in 0.13 seconds. There is a catch: those serving optimizations are not open source yet. NVIDIA expects the release-day vLLM path to be roughly 2x slower. On 1 H100 in BF16 with MTP2, NVIDIA estimates about 410 RTFx. For fine-tuning, NVIDIA tested 17 tasks. Each took about 0.5 to 2.3 GPU-hours on a single GB300. How does it compare with other open ASR models? *Competitor scores on the current 9-set leaderboard come from NVIDIA’s report, citing Hugging Face leaderboard results from October 2026. The leaderboard set changed since the competitors published their own cards, so the 2 WER rows are not interchangeable. Key Takeaways One open model handles transcripts, speakers, timestamps and translation across 40 languages. 4.71% English WER is the best open-weights score; commercial APIs still lead. About 4.3B active parameters; BF16 fits 1 H100, NVFP4 fits an RTX 5090. An entity list cuts named-entity WER from 12.21% to 2.20%. Offline only; Mandarin meetings and code-switching remain weak spots. Check out the technical report, model weights, project page and live demo. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post NVIDIA Releases Nemotron 3.5 Transcribe: An Open 30B MoE Model for 40-Language Transcription with ~4B Active Parameters appeared first on MarkTechPost.

Advertisement
AI

Google Deepmind releases Gemini 4 Carbon just two weeks after Argon with gains across the board

Just two weeks after Gemini 4 Argon, Google DeepMind is releasing Gemini 4 Carbon, which tops the Artificial Analysis Intelligence Index v5 with 59 points. Google says the model is built for extended agent workflows and the full cyber defense lifecycle. Because of those advanced security capabilities, access is limited to selected security experts at first. The article Google Deepmind releases Gemini 4 Carbon just two weeks after Argon with gains across the board appeared first on The Decoder.

AI

Google DeepMind Introduces Gemini 4 Carbon, Its Top Model for Cyber Defense and Coding

Google DeepMind has introduced Gemini 4 Carbon, its strongest model to date. Gemini 4 Carbon is a frontier model from Google built for long-horizon agentic coding and cybersecurity defense. It ranks #1 on the updated v5.0 Artificial Analysis Intelligence Index with a score of 59. For developers, the catch is access. Carbon starts with a vetted group of cyber defenders, not the public API. TL;DR Size: Parameter count and context window not disclosed. Its predecessor, Gemini 4 Argon, has a 1,048,576-token context. Runs on: Google-hosted only. No open weights, no self-hosting, no quantized variants. Performance: Leads 15 of 17 published benchmarks versus Argon, GPT-6 Astra, Opus 5.5 and Fable 5.1. Best: 96.2% BFS F1 on GraphWalks (256K to 1M tokens), versus 84.2% for Argon. Worst: 81.5% on OSWorld-2.1, trailing Claude Opus 5.5 at 87.2%. Bottom line: Best: state-of-the-art vulnerability finding, exploitation and patching. Worst: you probably cannot call it yet. What is Gemini 4 Carbon? Carbon is the second Gemini 4 model, following Gemini 4 Argon. Google team states that the model targets hard, long-running engineering work. Internal teams saw Carbon run for days on single tasks. Examples include root-causing and patching subtle bugs across large codebases and low-level systems. Google also tuned Carbon for full-stack web development and multimodal understanding. On WebDev Arena, Carbon gained 89 Elo points to reach #4 at 1769. How does Gemini 4 Carbon perform on benchmarks? Google published a 17-row comparison in its launch post. Carbon was run at the highest thinking setting. Most rival numbers are self-reported by each provider, per Google’s evaluation methodology. Agentic coding: Carbon scores 65.2% on Terminal-Bench 4.0 (AA), ahead of Opus 5.5 at 59.6%. It hits 9.6% on SWE-sweep, narrowly ahead of GPT-6 Astra at 9.5%. On FrontierSWE, GPT-6 Astra leads at 65.5% versus Carbon’s 61.2%. Computer use and knowledge work: Carbon reaches 77.4% on Agents’ Last Exam v1.1. On Vending-Bench 2, it ends with a $16,484 bank balance, the highest listed. Long context and science: GraphWalks is the widest gap: 96.2% versus 71.8% for GPT-6 Astra. Carbon also posts 80.0% on RiemannBench. Why does Carbon relevant for cybersecurity? Google frames Carbon around the full defense lifecycle: find, validate, then fix. Cyber benchmarks were evaluated without guardrails. Finding: 88.2% on SEC-bench Pro, versus 85.4% for GPT-6 Astra. Exploiting: 54.8% on ExploitGym, versus 42.4% for GPT-6 Astra and 34.5% for Opus 5.5. Patching: 71.0% on Collinear’s CWE-bench, versus 68.0% for Argon and GPT-6 Astra. Google computed SEC-bench Pro and ExploitGym scores itself, using its Antigravity harness. ExploitGym ran with mitigations off. Internally, Carbon showed more than 2x higher success at generating secure code than Argon. Google team states that the Chrome team found Carbon over 10x better than another lab’s model at high-quality patches. Wiz reports Carbon leading its web penetration testing benchmarks. <iframe id="mtp-g4c-frame" title="Gemini 4 Carbon explainer" srcdoc="<!doctype html> <title>Gemini 4 Carbon Explainer</title> html,body{margin:0!important;padding:0!important;background:transparent!important;} #mtp-g4c{--blue:#1A73E8;--b2:#4285F4;--red:#EA4335;--yel:#FBBC04;--grn:#34A853;--pur:#9B72CB;--ink:#1F1F1F;--mut:#5F6368;--line:#DADCE0;--soft:#F0F4F9; font-family:"Google Sans",Inter,Roboto,Arial,sans-serif!important;color:var(--ink)!important;background:#FFFFFF!important;border:1px solid var(--line)!important;border-radius:18px!important;max-width:880px;margin:0 auto;overflow:hidden;box-sizing:border-box;} #mtp-g4c *{box-sizing:border-box;} #mtp-g4c hr,#mtp-g4c p:empty,#mtp-g4c del,#mtp-g4c s{display:none!important;} #mtp-g4c .strip{height:5px!important;background:linear-gradient(90deg,var(--b2) 0 25%,var(--red) 25% 50%,var(--yel) 50% 75%,var(--grn) 75%)!important;} #mtp-g4c .head{padding:20px 24px 6px;} #mtp-g4c .eyebrow{font-size:12px;letter-spacing:.08em;text-transform:uppercase;color:var(--blue)!important;font-weight:700;} #mtp-g4c h3{margin:6px 0 4px!important;font-size:24px!important;line-height:1.2!important;color:var(--ink)!important;font-weight:700!important;border:0!important;padding:0!important;} #mtp-g4c h3 span{background:linear-gradient(90deg,var(--b2),var(--pur));-webkit-background-clip:text;background-clip:text;color:transparent!important;} #mtp-g4c .sub{margin:0!important;color:var(--mut)!important;font-size:14px;} #mtp-g4c .tabs{display:flex;gap:6px;padding:14px 24px 0;flex-wrap:wrap;border-bottom:1px solid var(--line)!important;} #mtp-g4c .tab{appearance:none;border:0!important;background:transparent!important;color:var(--mut)!important;font:inherit;font-size:14px;font-weight:600;padding:10px 12px;border-radius:10px 10px 0 0;cursor:pointer;position:relative;} #mtp-g4c .tab.on{color:var(--blue)!important;background:var(--soft)!important;} #mtp-g4c .tab.on::after{content:"";position:absolute;left:10px;right:10px;bottom:-1px;height:3px;border-radius:3px;background:var(--blue);} #mtp-g4c .panel{display:none;padding:20px 24px 22px;animation:g4cIn .35s ease;} #mtp-g4c .panel.on{display:block;} @keyframes g4cIn{from{opacity:0;transform:translateY(6px)}to{opacity:1;transform:none}} #mtp-g4c .grid4{display:grid;grid-template-columns:repeat(4,1fr);gap:10px;} #mtp-g4c .tile{background:var(--soft)!important;border:1px solid var(--line)!important;border-radius:12px;padding:12px;} #mtp-g4c .tile b{display:block;font-size:24px;color:var(--blue)!important;} #mtp-g4c .tile small{color:var(--mut)!important;font-size:12px;line-height:1.35;display:block;} #mtp-g4c .flow{display:flex;align-items:stretch;gap:8px;margin-top:16px;} #mtp-g4c .node{flex:1;border:1px solid var(--line)!important;border-radius:12px;padding:12px;background:#fff!important;opacity:.35;transform:scale(.97);transition:all .4s;} #mtp-g4c .node.lit{opacity:1;transform:none;border-color:var(--blue)!important;box-shadow:0 4px 14px rgba(26,115,232,.15);} #mtp-g4c .node h4{margin:0 0 4px!important;font-size:14px!important;color:var(--ink)!important;} #mtp-g4c .node p{margin:0!important;font-size:12.5px;color:var(--mut)!important;line-height:1.4;} #mtp-g4c .arrow{align-self:center;color:var(--blue)!important;font-weight:700;} #mtp-g4c .btns{display:flex;gap:8px;margin-top:14px;flex-wrap:wrap;} #mtp-g4c .btn{appearance:none;border:1px solid var(--blue)!important;background:#fff!important;color:var(--blue)!important;font:inherit;font-size:13px;font-weight:600;padding:8px 14px;border-radius:999px;cursor:pointer;} #mtp-g4c .btn.pri{background:var(--blue)!important;color:#fff!important;} #mtp-g4c .btn.on{background:var(--blue)!important;color:#fff!important;} #mtp-g4c .stage{display:grid;grid-template-columns:1fr 1.3fr;gap:16px;margin-top:6px;} #mtp-g4c .step{border:1px solid var(--line)!important;border-radius:12px;padding:12px 14px;margin-bottom:8px;cursor:pointer;background:#fff!important;transition:all .3s;} #mtp-g4c .step.on{border-color:var(--blue)!important;background:var(--soft)!important;} #mtp-g4c .step b{font-size:14px;color:var(--ink)!important;} #mtp-g4c .step small{display:block;color:var(--mut)!important;font-size:12px;margin-top:2px;} #mtp-g4c .bars{padding:4px 0;} #mtp-g4c .bt{font-size:13px;color:var(--mut)!important;margin:0 0 10px!important;} #mtp-g4c .row{display:grid;grid-template-columns:120px 1fr 64px;gap:8px;align-items:center;margin:7px 0;font-size:13px;} #mtp-g4c .track{height:14px;border-radius:7px;background:#EEF1F5!important;overflow:hidden;} #mtp-g4c .fill{height:100%;width:0;border-radius:7px;background:#BDC1C6!important;transition:width .9s cubic-bezier(.2,.8,.2,1);} #mtp-g4c .fill.me{background:linear-gradient(90deg,var(--b2),var(--pur))!important;} #mtp-g4c .val{font-weight:700;text-align:right;color:var(--ink)!important;} #mtp-g4c .nd{color:var(--mut)!important;font-weight:400;font-size:12px;} #mtp-g4c .chips{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:12px;} #mtp-g4c .note{font-size:11.5px;color:var(--mut)!important;margin-top:10px!important;line-height:1.45;} #mtp-g4c .sgrid{display:grid;grid-template-columns:repeat(3,1fr);gap:10px;} #mtp-g4c .scard{border:1px solid var(--line)!important;border-radius:12px;padding:14px;background:#fff!important;cursor:pointer;transition:all .25s;} #mtp-g4c .scard:hover{border-color:var(--grn)!important;transform:translateY(-2px);} #mtp-g4c .scard b{display:block;font-size:26px;color:var(--grn)!important;} #mtp-g4c .scard span{font-size:12.5px;color:var(--mut)!important;line-height:1.4;display:block;} #mtp-g4c .scard em{display:none;font-style:normal;font-size:12px;color:var(--ink)!important;margin-top:6px;} #mtp-g4c .scard.open em{display:block;} #mtp-g4c .tl{position:relative;margin:8px 0 4px;padding-left:22px;} #mtp-g4c .tl::before{content:"";position:absolute;left:7px;top:4px;bottom:4px;width:2px;background:var(--line);} #mtp-g4c .ti{position:relative;margin:0 0 12px;opacity:.4;transition:opacity .4s;} #mtp-g4c .ti.lit{opacity:1;} #mtp-g4c .ti::before{content:"";position:absolute;left:-20px;top:4px;width:12px;height:12px;border-radius:50%;background:#fff;border:2px solid var(--line);} #mtp-g4c .ti.lit::before{background:var(--blue);border-color:var(--blue);} #mtp-g4c .ti b{font-size:14px;color:var(--ink)!important;} #mtp-g4c .ti small{display:block;font-size:12.5px;color:var(--mut)!important;} #mtp-g4c .calc{background:var(--soft)!important;border:1px solid var(--line)!important;border-radius:12px;padding:14px;margin-top:8px;} #mtp-g4c .calc label{font-size:13px;color:var(--ink)!important;display:block;margin:8px 0 4px;} #mtp-g4c .calc input[type=range]{width:100%;accent-color:#1A73E8;} #mtp-g4c .cost{display:flex;gap:10px;margin-top:10px;} #mtp-g4c .cost div{flex:1;background:#fff!important;border:1px solid var(--line)!important;border-radius:10px;padding:10px;} #mtp-g4c .cost b{display:block;font-size:20px;color:var(--blue)!important;} #mtp-g4c .cost small{font-size:11.5px;color:var(--mut)!important;} #mtp-g4c .foot{display:flex;justify-content:space-between;gap:10px;flex-wrap:wrap;padding:10px 24px 14px;border-top:1px solid var(--line)!important;font-size:11.5px;color:var(--mut)!important;} #mtp-g4c .foot a{color:var(--blue)!important;text-decoration:none!important;} #mtp-g4c .brand{color:#76B900!important;font-weight:700;} @media (max-width:640px){ #mtp-g4c .grid4{grid-template-columns:repeat(2,1fr);} #mtp-g4c .flow{flex-direction:column;} #mtp-g4c .arrow{transform:rotate(90deg);} #mtp-g4c .stage{grid-template-columns:1fr;} #mtp-g4c .sgrid{grid-template-columns:1fr 1fr;} #mtp-g4c .row{grid-template-columns:92px 1fr 56px;font-size:12px;} #mtp-g4c h3{font-size:20px!important;} #mtp-g4c .head,#mtp-g4c .panel{padding-left:16px;padding-right:16px;} #mtp-g4c .tabs{padding:10px 12px 0;} #mtp-g4c .tab{font-size:12.5px;padding:8px 9px;} } @media (prefers-reduced-motion:reduce){#mtp-g4c *{transition:none!important;animation:none!important;}} <div id="mtp-g4c"> <div class="strip"></div> <div class="head"> <div class="eyebrow">Explainer · Google DeepMind · Oct 28, 2026</div> <h3>How <span>Gemini 4 Carbon</span> works and where it leads</h3> <p class="sub">Benchmarks from Google's launch post. Rival scores are mostly self-reported by each lab.</p> </div> <div class="tabs" role="tablist"> <button class="tab on" data-t="p1">Overview</button> <button class="tab" data-t="p2">Cyber lifecycle</button> <button class="tab" data-t="p3">Benchmarks</button> <button class="tab" data-t="p4">Safety</button> <button class="tab" data-t="p5">Access & price</button> </div> <div class="panel on" id="p1"> <div class="grid4"> <div class="tile"><b data-count="59">59</b><small>#1 on Artificial Analysis Index v5.0</small></div> <div class="tile"><b data-count="15">15</b><small>of 17 published benchmarks led</small></div> <div class="tile"><b>#4</b><small>WebDev Arena, 1769 Elo (+89)</small></div> <div class="tile"><b>$2/$10</b><small>per 1M tokens, introductory</small></div> </div> <div class="flow" id="g4cflow"> <div class="node"><h4>1 · Hard task in</h4><p>Long-horizon engineering work, like root-causing bugs in large codebases.</p></div> <div class="arrow">→</div> <div class="node"><h4>2 · Sustained agent loop</h4><p>Google says Carbon ran for days, pushing through dead ends.</p></div> <div class="arrow">→</div> <div class="node"><h4>3 · Safety stack</h4><p>Misuse defenses, prompt-injection resilience, chain-of-thought monitoring.</p></div> <div class="arrow">→</div> <div class="node"><h4>4 · Verified output</h4><p>Patches and code, with caveats surfaced instead of fabricated claims.</p></div> </div> <div class="btns"><button class="btn pri" id="g4cplay"> Play the loop</button></div> </div> <div class="panel" id="p2"> <div class="stage"> <div> <div class="step on" data-s="0"><b>1 · Find</b><small>SEC-bench Pro: uncover real vulnerabilities</small></div> <div class="step" data-s="1"><b>2 · Validate</b><small>ExploitGym: check which flaws are exploitable</small></div> <div class="step" data-s="2"><b>3 · Patch</b><small>CWE-bench: fix the root cause in code</small></div> <div class="btns"><button class="btn" id="g4cnext">Next step →</button></div> </div> <div class="bars" id="g4ccyber"></div> </div> <p class="note">Cyber benchmarks evaluated without guardrails. Google self-computed SEC-bench Pro and ExploitGym (mitigations off). Not disclosed = no score published.</p> </div> <div class="panel" id="p3"> <div class="chips" id="g4cchips"></div> <div class="bars" id="g4cbench"></div> <p class="note">Carbon and Argon run at highest thinking setting. Rivals at max available reasoning, mostly self-reported.</p> </div> <div class="panel" id="p4"> <div class="sgrid"> <div class="scard"><b>0.0%</b><span>Successful cheating on Terminal-Bench 4.0</span><em>Cheating attempts dropped to 0.4%, while accuracy improved.</em></div> <div class="scard"><b>~80%</b><span>Fewer high-severity misalignment alerts</span><em>Observed in live coding environments with Google's internal teams.</em></div> <div class="scard"><b>~2x</b><span>Fewer unsupported claims than Argon</span><em>Carbon writes caveats into the deliverable instead of hedging.</em></div> <div class="scard"><b>-26%</b><span>Overreach</span><em>Stays within scope on multi-step plans.</em></div> <div class="scard"><b>-29%</b><span>Under-service</span><em>Follows through on what was asked.</em></div> <div class="scard"><b>0.0%</b><span>Gray Swan IPI attack success</span><em>Argon was already at 0.7%. Carbon resisted every evaluated attack.</em></div> </div> <p class="note">Tap a card for detail. All figures are Google's internal or self-run results.</p> </div> <div class="panel" id="p5"> <div class="tl" id="g4ctl"> <div class="ti"><b>Pre-release testing</b><small>Google participates in the U.S. government's voluntary pre-release model testing.</small></div> <div class="ti"><b>This week: cyber defenders</b><small>A subset of cyber defenders gets access first.</small></div> <div class="ti"><b>Fairwind Program</b><small>Phased access for vetted partners: governments, critical infrastructure, core tech platforms.</small></div> <div class="ti"><b>Later: developers, businesses, consumers</b><small>No date announced. Google says "as soon as possible".</small></div> </div> <div class="calc"> <label>Monthly input tokens: <b id="g4cin">50</b>M</label> <label>Monthly output tokens: <b id="g4cout">10</b>M</label> <div class="cost"> <div><b id="g4cc1">$200</b><small>Intro price ($2 / $10)</small></div> <div><b id="g4cc2">$400</b><small>After intro ($4 / $20)</small></div> </div> </div> <p class="note">Estimate excludes caching. Cached input is 95% off the input price.</p> </div> <div class="foot"> <span>Source: <a href=" target=" rel="noopener">Google launch post</a> · <a href=" target=" rel="noopener">methodology</a></span> <span class="brand">© Marktechpost</span> </div> </div> (function(){ var R=document.getElementById('mtp-g4c'); function post(){try{parent.postMessage({mtpEmbed:'g4c',height:R.offsetHeight+40},'*');}catch(e){}} var M=['Gemini 4 Carbon','Gemini 4 Argon','GPT-6 Astra','Claude Opus 5.5','Claude Fable 5.1']; function bars(el,title,vals,max,unit){ var h='<p class="bt">'+title+'</p>'; vals.forEach(function(v,i){ h+='<div class="row"><span>'+M[i]+'</span><div class="track"><div class="fill&apos;+(i==0?&apos; me&apos;:&apos;&apos;)+&apos;" data-w="&apos;+(v==null?0:v/max*100)+&apos;"></div></div><span class="val">'+(v==null?'<span class="nd">Not disclosed</span>':(unit=='$'?'$'+v.toLocaleString('en-US'):v+unit))+'</span></div>'; }); el.innerHTML=h; setTimeout(function(){el.querySelectorAll('.fill').forEach(function(f){f.style.width=f.getAttribute('data-w')+'%';});},60); post(); } var CY=[['SEC-bench Pro (find)',[88.2,83.4,85.4,null,null]],['ExploitGym (validate)',[54.8,42.0,42.4,34.5,null]],['CWE-bench (patch)',[71.0,68.0,68.0,67.0,58.0]]]; var cs=0,cyEl=document.getElementById('g4ccyber'); function cy(i){cs=i;document.querySelectorAll('#mtp-g4c .step').forEach(function(s,j){s.classList.toggle('on',j==i);});bars(cyEl,CY[i][0],CY[i][1],100,'%');} document.querySelectorAll('#mtp-g4c .step').forEach(function(s,j){s.onclick=function(){cy(j);};}); document.getElementById('g4cnext').onclick=function(){cy((cs+1)%3);}; var B=[ ['AA Index v5.0',[59,50,57,58,49],70,''], ['Terminal-Bench 4.0 (AA)',[65.2,57.1,59.1,59.6,52.0],100,'%'], ['FrontierSWE',[61.2,55.0,65.5,62.3,56.3],100,'%'], ['OSWorld-2.1 (offline subset)',[81.5,74.7,80.9,87.2,76.8],100,'%'], ['Agents’ Last Exam v1.1',[77.4,69.0,73.1,71.6,null],100,'%'], ['GraphWalks 256K to 1M (BFS F1)',[96.2,84.2,71.8,66.8,65.0],100,'%'], ['Vending-Bench 2 (bank balance)',[16484,13718,15515,9235,5422],18000,'$'], ['RiemannBench',[80.0,76.0,72.0,69.6,65.6],100,'%'] ]; var ch=document.getElementById('g4cchips'),be=document.getElementById('g4cbench'); B.forEach(function(b,i){var x=document.createElement('button');x.className='btn'+(i==0?' on':'');x.textContent=b[0];x.onclick=function(){ch.querySelectorAll('.btn').forEach(function(y){y.classList.remove('on');});x.classList.add('on');bars(be,b[0],b[1],b[2],b[3]);};ch.appendChild(x);}); function flow(){var n=document.querySelectorAll('#g4cflow .node');n.forEach(function(x){x.classList.remove('lit');});n.forEach(function(x,i){setTimeout(function(){x.classList.add('lit');},i*550);});} document.getElementById('g4cplay').onclick=flow; function tl(){var t=document.querySelectorAll('#g4ctl .ti');t.forEach(function(x){x.classList.remove('lit');});t.forEach(function(x,i){setTimeout(function(){x.classList.add('lit');},i*450);});} document.querySelectorAll('#mtp-g4c .scard').forEach(function(c){c.onclick=function(){c.classList.toggle('open');post();};}); var ri=document.getElementById('g4crin'),ro=document.getElementById('g4crout'); function calc(){var a=+ri.value,b=+ro.value;document.getElementById('g4cin').textContent=a;document.getElementById('g4cout').textContent=b; document.getElementById('g4cc1').textContent='$'+(a*2+b*10).toLocaleString('en-US');document.getElementById('g4cc2').textContent='$'+(a*4+b*20).toLocaleString('en-US');} ri.oninput=calc;ro.oninput=calc;calc(); var started={}; document.querySelectorAll('#mtp-g4c .tab').forEach(function(t){t.onclick=function(){ document.querySelectorAll('#mtp-g4c .tab').forEach(function(x){x.classList.toggle('on',x==t);}); var id=t.getAttribute('data-t'); document.querySelectorAll('#mtp-g4c .panel').forEach(function(p){p.classList.toggle('on',p.id==id);}); if(id=='p1')flow(); if(id=='p2')cy(cs); if(id=='p3'){var on=ch.querySelector('.btn.on');on&&on.click();} if(id=='p5')tl(); post(); };}); cy(0);bars(be,B[0][0],B[0][1],B[0][2],B[0][3]); flow(); window.addEventListener('load',post);window.addEventListener('resize',post);setTimeout(post,300); })(); " style="width:100%;max-width:880px;height:600px;border:0;display:block;margin:24px auto;" loading="lazy"> window.addEventListener("message",function(e){var d=e.data;if(d&&d.mtpEmbed==="g4c"&&d.height){var f=document.getElementById("mtp-g4c-frame");if(f)f.style.height=d.height+"px";}}); What safety changes ship with Carbon? Carbon keeps Argon’s layered stack: training, real-time misuse defenses, prompt-injection resilience and chain-of-thought monitoring. Google reports several alignment gains: Successful cheating on Terminal-Bench 4.0 fell to 0.0%, with cheating attempts at 0.4%. High-severity misalignment alerts in live internal coding dropped by almost 80%. Unsupported claims fell roughly 2x versus Argon. Overreach fell 26% and under-service fell 29%. Attack success on Gray Swan’s Indirect Prompt Injection benchmark is 0.0%, down from Argon’s 0.7%. These are Google’s internal or self-run figures. Independent replication is not yet available. How can developers access Gemini 4 Carbon? Not yet, for most. Google is participating in the U.S. government’s voluntary pre-release testing process. This week, a subset of cyber defenders gets access. More trusted partners follow through the Fairwind Program, which prioritizes governments, critical infrastructure operators and core tech platforms. Developers, businesses and consumers get Carbon later, with no date given. Pricing matches Argon’s introductory rate: $2 per million input tokens and $10 per million output tokens. Cached input is 95% off. After the introductory period, prices rise to $4 and $20. Gemini 4 Carbon vs competitors Benchmark rows: Google launch post; rival scores mostly self-reported per methodology. Key Takeaways Gemini 4 Carbon ranks #1 on Artificial Analysis Index v5.0 at 59. It leads SEC-bench Pro (88.2%), ExploitGym (54.8%) and CWE-bench (71.0%). GraphWalks long context hits 96.2%, its widest margin. Claude Opus 5.5 still wins OSWorld-2.1; GPT-6 Astra wins FrontierSWE. Access starts with cyber defenders at $2/$10 per million tokens. Check out the Google’s launch post, the evaluation methodology and the Fairwind Program page. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Google DeepMind Introduces Gemini 4 Carbon, Its Top Model for Cyber Defense and Coding appeared first on MarkTechPost.

AI

MiniMax Open-Sources MCode Stack: 15 Agent Skills Its Coding Team Uses Daily

MiniMax has open-sourced MCode Stack, the workflows and agent skills its team uses to build MiniMax Code. The MiniMax MCode Stack agent skills went public on October 28, 2026 under the MIT license. It is not a prompt collection. Each piece is a SKILL.md skill or checklist that installs directly into MiniMax Code. For developers, it turns harness engineering habits into steps an agent can actually run. TL;DR Size: 2 stacks, 17 skills (15 from Ronny, 2 from Jiachen), 5 checklists, 5 companion notes. Runs on: Any machine running MiniMax Code (Node.js 22.19+) or Claude Code. No GPU. Skills are plain Markdown. Performance: No benchmarks or eval results published. Best: long-command-check, a 10-point audit for long shell commands and background tasks. What is MiniMax MCode Stack? MCode Stack is an open-source repo of agent skills from MiniMax’s MiniMax Code team. MiniMax Code is the MiniMax’s open-source terminal coding agent, also MIT-licensed. The stack holds no MiniMax Code source code. It holds only workflows, skills and checklists. The repo pairs with a column called MCode Engineering Notes. Each note explains one idea. The repo makes that idea runnable. Published notes cover subagents, tool context, compaction versus memory and long bash commands. A note on per-turn context tax is marked upcoming. What skills are inside Ronny’s stack? MiniMax’s stack ships 15 skills, grouped by moment in the working day: Plan and split (3): subagent-split-check, engineering-small-prs, engineering-change-descriptions. Context and tools (2): tool-budget-audit, context-tax-audit. Long sessions and commands (3): long-session-check, handoff-doc, long-command-check. Review (4): independent-review, engineering-code-review, engineering-review-comments, engineering-review-feedback. Experiments and hotfixes (3): harness-experiment, trajectory-review, engineering-emergency-changes. A few ideas stand out. independent-review gives the reviewer exactly 3 inputs: the diff, the requirement and test results. It deliberately withholds the author’s conversation. The reasoning: a reviewer inheriting the author’s context repeats the author’s blind spots. subagent-split-check says to split work only when context is the measured problem. Splitting by team or question type is not a valid reason. tool-budget-audit separates 2 costs: tool definitions and tool results. Tool search mainly cuts the first. Programmatic tool calling mainly cuts the second. harness-experiment enforces 1 change per experiment. It tracks flipped tasks, requests per task, tokens, cache hit rate and wall-clock time. The long-command checklist has 10 items. It asks whether moving a command to background re-runs the process. It also asks whether unread results survive compaction. The 6 engineering-* skills are adapted from Google Engineering Practices. That material stays under CC-BY 3.0, per the stack’s NOTICE. window.addEventListener("message",function(e){var d=e.data;if(d&&d.id==="mtp-mcs"&&d.mtpEmbedHeight){var f=document.getElementById("mtp-mcs-frame");if(f)f.style.height=d.mtpEmbedHeight+"px";}}); How do you install MCode Stack? There are 3 paths, per the README: One skill: clone the repo and copy a skill folder into ~/.agents/skills/. Whole stack as a local plugin: copy stacks/ronny into MiniMax Code’s local plugin directory. Marketplace: in MiniMax Code desktop, add the MiniMax-AI/mcode-stack marketplace. This option is rolling out gradually. The same catalog works in Claude Code with /plugin marketplace add MiniMax-AI/mcode-stack. MiniMax Code also reads ~/.claude/skills/ and ~/.codex/skills/. The skills use the common SKILL.md format, so most work in other agents that support skills. How does MCode Stack compare with other skill packs? MCode Stack is narrower than both. It targets people who design agent harnesses, not only people who use agents. Superpowers covers far more harnesses today. Key Takeaways MiniMax open-sourced 17 SKILL.md skills and 5 checklists under MIT. Skills pair with MCode Engineering Notes, turning essays into runnable checks. independent-review feeds a reviewer only the diff, requirement and test results. Installs into MiniMax Code or Claude Code via a plugin marketplace. No published evals yet, so treat it as a playbook, not proof. Check out the GitHub repo and MiniMax Code. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post MiniMax Open-Sources MCode Stack: 15 Agent Skills Its Coding Team Uses Daily appeared first on MarkTechPost.

Advertisement
AI

Google opens Android development to third-party AI agents and unveils new Gemini APIs

Google is opening up Android Studio to third-party AI agents. Through a new "Bring Your Own Agent" feature, the IDE now supports more than 40 coding agents, including Anthropic's Claude Agent and OpenAI's Codex. Google says new terminal tools also help agents finish UI tasks 77 percent faster while using up to three times fewer tokens. Android development is becoming agent-centric too. The article Google opens Android development to third-party AI agents and unveils new Gemini APIs appeared first on The Decoder.

AI

OpenAI teases Dots for Europe, but its Berlin DevDay was really about building a platform

At DevDay Exchange in Berlin, OpenAI mostly repeated what it had already announced, but it did share a few new details. The AI assistant Dots should roll out more widely in Europe "in a few weeks," and weekly Codex users in Germany have grown twentyfold since January. A clear subtext ran through the sessions and the repeated nods to ChatGPT's 1.2 billion weekly users: OpenAI wants developers to build their businesses directly on its infrastructure. The article OpenAI teases Dots for Europe, but its Berlin DevDay was really about building a platform appeared first on The Decoder.

AI

Claude for Legal now verifies sources automatically, then hands the file to a lawyer

Anthropic is adding pre-built templates to Claude for Legal. These interactive overviews and tables simplify setup for legal teams and speed up routine work like contract reviews, while more than 60 integrations let firms connect their own databases directly. The article Claude for Legal now verifies sources automatically, then hands the file to a lawyer appeared first on The Decoder.

Advertisement
AI

Nvidia’s open Nemotron 3.5 Transcribe combines transcription, translation, and speaker recognition

Nvidia has released Nemotron 3.5 Transcribe, an open speech recognition model that handles transcription, speaker recognition, and translation and takes its instructions as text prompts. It leads several benchmarks and uses only a sixth of the active parameters of its strongest open rival. Nvidia is also shipping a dataset with more than 800,000 hours of speech, along with tools for fine-tuning. The article Nvidia's open Nemotron 3.5 Transcribe combines transcription, translation, and speaker recognition appeared first on The Decoder.