Nanbeige LLM Lab has released Nanbeige4.3-33B on Hugging Face under the Apache 2.0 license. Nanbeige4.3-33B is a 33B dense language model built for agentic coding and office-style ‘cowork’ tasks. The lab sits inside the recruitment platform BOSS Zhipin. It also published Nanbeige4.3-33B-Base for custom post-training. The model supports a 524,288-token context window. In the lab’s own table, it leads 17 of 20 benchmarks against 3 open models near 30B. TL;DR Size: 33B parameters, dense (no MoE), 524,288-token (512K) context, BF16 weights. Runs on: Multi-GPU servers. The card’s SGLang and vLLM recipes use 4-way tensor parallelism. BF16 weights need about 66 GB (our arithmetic). 1 community quantization is listed. Performance: Leads 17 of 20 rows in Nanbeige’s own comparison against Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B. Best: 80.9 on Terminal-Bench 2.1, versus 73.0 for Qwen3.8-27B. Worst: 23.8 on Terminal-Bench 4.0 (its lowest score). It also trails Qwen3.8-27B on SWE-Bench Pro, 60.3 vs 61.7. What is Nanbeige4.3-33B? Nanbeige4.3-33B is the flagship ‘teacher’ model of the Nanbeige4.3 series. The lab’s stated strategy is to build a strong teacher first. It then plans to distill agentic skills into smaller models, starting with Nanbeige4.3-3B. How is Nanbeige4.3-33B built? The config.json declares a standard LlamaForCausalLM architecture. There is no expert routing, so all 33B parameters run on every token. 52 layers, hidden size 7,168 56 query heads and 8 key-value heads (grouped-query attention), head dimension 128 FFN intermediate size 22,016, vocabulary of 166,912 tokens RoPE theta of 120,000,000 and 524,288 max positions The chat template exposes 3 controls. enable_thinking toggles reasoning and is on by default. preserve_thinking keeps reasoning from earlier turns. Nanbeige recommends it for multi-turn tool use and code agents. Passing tools enables tool calling, and an XML call format is recommended. How does Nanbeige4.3-33B perform on benchmarks? Nanbeige compared the model with Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B. Every model ran at its highest thinking effort, with shared hyperparameters and harnesses. All scores in this section are self-reported. Agentic coding: It scores 80.9 on Terminal-Bench 2.1, against 73.0 for Qwen3.8-27B. DeepSWE shows the widest gap: 65.5 versus 42.5. On Terminal-Bench 4.0, it scores 23.8 while Gemma4-31B scores 0.0. Agentic cowork: It posts 54.2 on Automation Bench versus 38.3, and 44.8 on JobBench versus 34.5. On YC-bench, simulated earnings reach $1,025,437 versus $641,290. Reasoning: It reports 95.2 on AIME 2026 and 84.0 on AA-LCR. It loses 3 of 20 rows. Qwen3.8-27B leads SWE-Bench Pro (61.7 vs 60.3) and GDPval (86.4 vs 84.8). Gemma4-31B leads GPQA Diamond in Nanbeige’s table, 91.3 vs 89.5. The Base model beats Gemma4-31B-Base on all 8 reported tasks. Examples include MMLU-Pro (66.9 vs 59.3) and GPQA (62.0 vs 41.5). <iframe id="mtp-nbg43-frame" title="Nanbeige4.3-33B interactive explainer" srcdoc="<!doctype html> <title>Nanbeige4.3-33B Explainer</title> html,body{margin:0;padding:0;background:transparent} #mtp-nbg43{--bg:#0b1d1c;--card:#112b29;--line:#24514b;--mint:#5fd3bd;--mint2:#8ee6d2;--ink:#e8f6f2;--mute:#9cbcb5;--qwen:#7c8cf8;--muse:#f0a35e;--gemma:#6aa9f4;--mtp:#76B900; font-family:Inter,-apple-system,Segoe UI,Roboto,sans-serif!important;background:var(--bg)!important;color:var(--ink)!important;border:1px solid var(--line)!important;border-radius:16px;max-width:880px;margin:0 auto;padding:22px 24px 14px;box-sizing:border-box;overflow:hidden} #mtp-nbg43 *{box-sizing:border-box} #mtp-nbg43 hr,#mtp-nbg43 p:empty,#mtp-nbg43 del,#mtp-nbg43 s{display:none!important} #mtp-nbg43 .hd{display:flex;align-items:center;justify-content:space-between;gap:12px;flex-wrap:wrap} #mtp-nbg43 .tag{font-size:11px;letter-spacing:.12em;text-transform:uppercase;color:var(--mint)!important;font-weight:700} #mtp-nbg43 h3{margin:4px 0 2px;font-size:22px;color:var(--ink)!important;font-weight:800;line-height:1.2} #mtp-nbg43 .sub{color:var(--mute)!important;font-size:13px;margin:0} #mtp-nbg43 .chips{display:flex;gap:6px;flex-wrap:wrap} #mtp-nbg43 .chip{font-size:11px;padding:4px 9px;border-radius:999px;border:1px solid var(--line)!important;color:var(--mint2)!important;background:#0e2523!important;font-weight:600} #mtp-nbg43 .tabs{display:flex;gap:6px;margin:18px 0 14px;border-bottom:1px solid var(--line)!important;flex-wrap:wrap} #mtp-nbg43 .tab{background:transparent!important;border:0!important;color:var(--mute)!important;font:600 13px Inter,sans-serif;padding:9px 12px;cursor:pointer;border-bottom:2px solid transparent!important;margin-bottom:-1px;transition:color .2s} #mtp-nbg43 .tab.on{color:var(--ink)!important;border-bottom-color:var(--mint)!important} #mtp-nbg43 .pane{display:none;animation:nbgIn .35s ease} #mtp-nbg43 .pane.on{display:block} @keyframes nbgIn{from{opacity:0;transform:translateY(6px)}to{opacity:1;transform:none}} /* flow */ #mtp-nbg43 .flow{display:grid;grid-template-columns:repeat(4,1fr);gap:10px;position:relative} #mtp-nbg43 .node{background:var(--card)!important;border:1px solid var(--line)!important;border-radius:12px;padding:12px;min-height:96px;transition:all .35s;opacity:.45;cursor:pointer} #mtp-nbg43 .node.on{opacity:1;border-color:var(--mint)!important;box-shadow:0 0 0 1px var(--mint),0 0 22px rgba(95,211,189,.25)} #mtp-nbg43 .node .n{font-size:11px;color:var(--mint)!important;font-weight:700} #mtp-nbg43 .node .t{font-weight:700;font-size:14px;margin:4px 0;color:var(--ink)!important} #mtp-nbg43 .node .d{font-size:12px;color:var(--mute)!important;line-height:1.4} #mtp-nbg43 .rail{height:4px!important;background:var(--line)!important;border-radius:4px;margin:12px 0 0;overflow:hidden} #mtp-nbg43 .rail i{display:block;height:100%;width:25%;background:linear-gradient(90deg,var(--mint),var(--mint2));transition:width .5s ease} #mtp-nbg43 .detail{background:#0e2523!important;border:1px solid var(--line)!important;border-radius:12px;padding:14px 16px;margin-top:12px;font-size:13.5px;line-height:1.55;min-height:92px} #mtp-nbg43 .detail b{color:var(--mint2)!important} #mtp-nbg43 .ctl{display:flex;gap:8px;margin-top:12px;align-items:center} #mtp-nbg43 .btn{background:var(--mint)!important;color:#06201c!important;border:0!important;border-radius:8px;font:700 13px Inter,sans-serif;padding:8px 14px;cursor:pointer} #mtp-nbg43 .btn.ghost{background:transparent!important;color:var(--mint2)!important;border:1px solid var(--line)!important} #mtp-nbg43 .spec{display:grid;grid-template-columns:repeat(4,1fr);gap:8px;margin-top:14px} #mtp-nbg43 .spec div{background:var(--card)!important;border:1px solid var(--line)!important;border-radius:10px;padding:10px} #mtp-nbg43 .spec b{display:block;font-size:18px;color:var(--mint2)!important} #mtp-nbg43 .spec span{font-size:11px;color:var(--mute)!important} /* bars */ #mtp-nbg43 .seg{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:12px} #mtp-nbg43 .seg button{background:#0e2523!important;border:1px solid var(--line)!important;color:var(--mute)!important;font:600 12px Inter,sans-serif;padding:6px 11px;border-radius:999px;cursor:pointer} #mtp-nbg43 .seg button.on{background:var(--mint)!important;color:#06201c!important;border-color:var(--mint)!important} #mtp-nbg43 .legend{display:flex;gap:14px;flex-wrap:wrap;font-size:12px;color:var(--mute)!important;margin-bottom:10px} #mtp-nbg43 .legend i{display:inline-block;width:10px;height:10px;border-radius:3px;margin-right:5px;vertical-align:-1px} #mtp-nbg43 .bench{margin-bottom:12px} #mtp-nbg43 .bench .bn{font-size:13px;font-weight:700;margin-bottom:5px;color:var(--ink)!important} #mtp-nbg43 .brow{display:grid;grid-template-columns:110px 1fr 54px;gap:8px;align-items:center;font-size:11.5px;margin:2px 0} #mtp-nbg43 .brow .lb{color:var(--mute)!important;white-space:nowrap;overflow:hidden;text-overflow:ellipsis} #mtp-nbg43 .track{height:12px!important;background:#0e2523!important;border-radius:6px;overflow:hidden} #mtp-nbg43 .fill{height:100%;width:0;border-radius:6px;transition:width .9s cubic-bezier(.2,.8,.2,1)} #mtp-nbg43 .brow .v{text-align:right;font-variant-numeric:tabular-nums;color:var(--ink)!important} #mtp-nbg43 .brow.me .v,#mtp-nbg43 .brow.me .lb{color:var(--mint2)!important;font-weight:700} #mtp-nbg43 .note{font-size:11.5px;color:var(--mute)!important;margin-top:6px;line-height:1.45} /* calc */ #mtp-nbg43 .calc{display:grid;grid-template-columns:1.1fr 1fr;gap:16px} #mtp-nbg43 .field{margin-bottom:14px} #mtp-nbg43 .field label{display:flex;justify-content:space-between;font-size:12.5px;color:var(--mute)!important;margin-bottom:6px} #mtp-nbg43 .field label b{color:var(--mint2)!important} #mtp-nbg43 input[type=range]{width:100%;accent-color:#5fd3bd} #mtp-nbg43 .stack{background:var(--card)!important;border:1px solid var(--line)!important;border-radius:12px;padding:14px} #mtp-nbg43 .stackbar{display:flex;height:22px!important;border-radius:6px;overflow:hidden;background:#0e2523!important;margin:8px 0 10px} #mtp-nbg43 .stackbar i{display:block;height:100%;transition:width .5s ease} #mtp-nbg43 .total{font-size:30px;font-weight:800;color:var(--ink)!important} #mtp-nbg43 .kv{display:flex;justify-content:space-between;font-size:12.5px;margin:4px 0;color:var(--mute)!important} #mtp-nbg43 .kv b{color:var(--ink)!important} /* caveats */ #mtp-nbg43 table{width:100%;border-collapse:collapse;font-size:12.5px} #mtp-nbg43 th,#mtp-nbg43 td{border-bottom:1px solid var(--line)!important;padding:8px 6px;text-align:left;color:var(--ink)!important;background:transparent!important} #mtp-nbg43 th{color:var(--mint)!important;font-size:11px;text-transform:uppercase;letter-spacing:.06em} #mtp-nbg43 td.bad{color:#f6b26b!important;font-weight:700} #mtp-nbg43 .flag{display:inline-block;font-size:10.5px;padding:2px 7px;border-radius:999px;background:#3a2a12!important;color:#f6b26b!important;font-weight:700} #mtp-nbg43 .ft{display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:8px;border-top:1px solid var(--line)!important;margin-top:16px;padding-top:10px;font-size:11px;color:var(--mute)!important} #mtp-nbg43 .ft a{color:var(--mute)!important;text-decoration:underline} #mtp-nbg43 .brand{color:#76B900!important;font-weight:700} #mtp-nbg43 pre,#mtp-nbg43 code{background:#0e2523!important;color:var(--mint2)!important;border:0!important;font-family:ui-monospace,Menlo,monospace;font-size:12px;padding:1px 5px;border-radius:4px} @media (max-width:640px){ #mtp-nbg43{padding:16px 14px 10px} #mtp-nbg43 h3{font-size:18px} #mtp-nbg43 .flow,#mtp-nbg43 .spec{grid-template-columns:1fr 1fr} #mtp-nbg43 .calc{grid-template-columns:1fr} #mtp-nbg43 .brow{grid-template-columns:86px 1fr 48px} #mtp-nbg43 .tab{padding:8px 8px;font-size:12px} } <div id="mtp-nbg43"> <div class="hd"> <div> <div class="tag">Nanbeige LLM Lab · Open weights</div> <h3>Nanbeige4.3-33B, explained</h3> <p class="sub">A 33B dense teacher model for agentic coding and cowork tasks</p> </div> <div class="chips"><span class="chip">33B dense</span><span class="chip">512K context</span><span class="chip">Apache 2.0</span><span class="chip">Text-only</span></div> </div> <div class="tabs" role="tablist"> <button class="tab on" data-p="p1">How it works</button> <button class="tab" data-p="p2">Benchmarks</button> <button class="tab" data-p="p3">Memory math</button> <button class="tab" data-p="p4">Caveats</button> </div> <div class="pane on" id="p1"> <div class="flow"> <div class="node on" data-s="0"><div class="n">STEP 1</div><div class="t">Base model</div><div class="d">Nanbeige4.3-33B-Base, released for custom post-training</div></div> <div class="node" data-s="1"><div class="n">STEP 2</div><div class="t">Teacher agent</div><div class="d">Post-trained Nanbeige4.3-33B with thinking and tools</div></div> <div class="node" data-s="2"><div class="n">STEP 3</div><div class="t">Agent harness</div><div class="d">Runs inside harnesses like Terminus 2, mini-swe-agent, Claude Code</div></div> <div class="node" data-s="3"><div class="n">STEP 4</div><div class="t">Distillation</div><div class="d">Planned: agentic skills move into Nanbeige4.3-3B</div></div> </div> <div class="rail"><i id="nbgRail"></i></div> <div class="detail" id="nbgDetail"></div> <div class="ctl"><button class="btn ghost" id="nbgPrev"> Back</button><button class="btn" id="nbgNext">Next step </button><button class="btn ghost" id="nbgPlay">Auto-play</button></div> <div class="spec"> <div><b>52</b><span>layers</span></div> <div><b>56 / 8</b><span>query / KV heads (GQA)</span></div> <div><b>7,168</b><span>hidden size</span></div> <div><b>524,288</b><span>max positions</span></div> </div> </div> <div class="pane" id="p2"> <div class="seg" id="nbgSeg"> <button class="on" data-c="coding">Agentic coding</button> <button data-c="cowork">Agentic cowork</button> <button data-c="reason">Reasoning</button> <button data-c="base">Base model</button> </div> <div class="legend" id="nbgLegend"></div> <div id="nbgBars"></div> <p class="note">All scores self-reported by Nanbeige on its Hugging Face model cards, at each model's highest thinking effort with shared harnesses. Not independently verified.</p> </div> <div class="pane" id="p3"> <div class="calc"> <div> <div class="field"><label>Context length <b id="nbgCtxL">131,072 tokens</b></label></div> <div class="field"><label>Concurrent sequences <b id="nbgSeqL">1</b></label></div> <div class="seg" id="nbgW"><button class="on" data-w="2">Weights BF16</button><button data-w="1">Weights 8-bit</button></div> <div class="seg" id="nbgK"><button class="on" data-k="2">KV cache BF16</button><button data-k="1">KV cache FP8</button></div> <p class="note">Arithmetic from config.json: 52 layers × 8 KV heads × 128 dims × 2 (K and V) × bytes per value. That is 212,992 bytes (about 208 KiB) per token in BF16. Weights use 33B parameters. Excludes activations and framework overhead.</p> </div> <div class="stack"> <div class="kv"><span>Estimated GPU memory</span></div> <div class="total" id="nbgTot">0 GB</div> <div class="stackbar"><i id="nbgBw" style="background:#5fd3bd"></i><i id="nbgBk" style="background:#7c8cf8"></i></div> <div class="kv"><span><i style="display:inline-block;width:9px;height:9px;background:#5fd3bd;border-radius:2px"></i> Weights</span><b id="nbgWv"></b></div> <div class="kv"><span><i style="display:inline-block;width:9px;height:9px;background:#7c8cf8;border-radius:2px"></i> KV cache</span><b id="nbgKv"></b></div> <div class="kv"><span>Per 80 GB GPU (ideal fit)</span><b id="nbgGpu"></b></div> <p class="note">The model card's SGLang and vLLM recipes use 4-way tensor parallelism.</p> </div> </div> </div> <div class="pane" id="p4"> <table> <tr><th>Row in Nanbeige table</th><th>Nanbeige lists</th><th>Official card</th><th></th></tr> <tr><td>Muse-Glimmer-30B, SWE-Bench Pro</td><td class="bad">36.9</td><td>51.2 (Meta)</td><td><span class="flag">looks swapped</span></td></tr> <tr><td>Gemma4-31B, SWE-Bench Pro</td><td class="bad">51.2</td><td>36.9 (Meta-reported)</td><td><span class="flag">looks swapped</span></td></tr> <tr><td>Gemma4-31B, GPQA Diamond</td><td class="bad">91.3</td><td>84.3 (Google)</td><td><span class="flag">conflict</span></td></tr> <tr><td>Muse-Glimmer-30B, Terminal-Bench 2.1</td><td class="bad">61.8</td><td>51.7 (Meta)</td><td><span class="flag">harness?</span></td></tr> <tr><td>Qwen3.8-27B, GPQA Diamond</td><td>89.2</td><td>89.2 (Qwen)</td><td>match</td></tr> <tr><td>Muse-Glimmer-30B, AIME 2026</td><td>94.7</td><td>94.7 (Meta)</td><td>match</td></tr> </table> <p class="note">Nanbeige re-ran every model with its own harness, which explains some gaps. Other limits: text-only input, dense compute on every token, and Nanbeige4.3-3B is not yet released.</p> </div> <div class="ft"><span>Sources: <a href=" target=" rel="noopener">Nanbeige model card</a> · <a href=" target=" rel="noopener">Base card</a> · competitor HF cards</span><span class="brand">© Marktechpost</span></div> </div> (function(){ var R=document.getElementById('mtp-nbg43'); function resize(){try{parent.postMessage({type:'mtp-nbg43-resize',height:R.offsetHeight+40},'*')}catch(e){}} // tabs 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.remove('on')}); document.getElementById(t.dataset.p).classList.add('on'); if(t.dataset.p==='p2')drawBars(cur); setTimeout(resize,50);setTimeout(resize,400); })}); // flow var steps=[ '<b>Base model.</b> Nanbeige4.3-33B-Base is a dense LlamaForCausalLM checkpoint. Nanbeige reports it beats Gemma4-31B-Base on all 8 listed tasks, for example MMLU-Pro 66.9 vs 59.3.', '<b>Teacher agent.</b> The post-trained model thinks by default (<code>enable_thinking</code>). <code>preserve_thinking</code> keeps earlier reasoning for multi-turn tool use. Nanbeige recommends the XML tool-call format.', '<b>Agent harness.</b> Nanbeige evaluated it in Terminus 2, mini-swe-agent v2.4.6 and the Claude Code harness. Recommended agent sampling: temperature 1.0, top_p 0.95, 65,536 max new tokens.', '<b>Distillation.</b> The 33B model is the teacher for the Nanbeige4.3 series. The first planned student is Nanbeige4.3-3B. No release date is given.' ]; var s=0,timer=null; function setStep(i){s=(i+4)%4;R.querySelectorAll('.node').forEach(function(n,k){n.classList.toggle('on',k<=s)}); document.getElementById('nbgRail').style.width=((s+1)*25)+'%';document.getElementById('nbgDetail').innerHTML=steps[s];resize()} R.querySelectorAll('.node').forEach(function(n){n.addEventListener('click',function(){stop();setStep(+n.dataset.s)})}); function stop(){if(timer){clearInterval(timer);timer=null;document.getElementById('nbgPlay').textContent='Auto-play'}} document.getElementById('nbgNext').onclick=function(){stop();setStep(s+1)}; document.getElementById('nbgPrev').onclick=function(){stop();setStep(s-1)}; document.getElementById('nbgPlay').onclick=function(){if(timer){stop();return}this.textContent='Pause';timer=setInterval(function(){setStep(s+1)},2200)}; setStep(0); // bars var M={n:['Nanbeige4.3-33B','#5fd3bd'],q:['Qwen3.8-27B','#7c8cf8'],m:['Muse-Glimmer-30B','#f0a35e'],g:['Gemma4-31B','#6aa9f4']}; var D={ coding:[['Terminal-Bench 2.1',80.9,73.0,61.8,47.2],['DeepSWE',65.5,42.5,4.4,2.7],['SWE-Bench Pro',60.3,61.7,36.9,51.2],['SkillsBench',61.1,56.4,39.7,28.9],['Terminal-Bench 4.0',23.8,6.3,1.6,0.0]], cowork:[['Automation Bench',54.2,38.3,15.2,9.7],['JobBench',44.8,34.5,16.6,9.0],['GDPval',84.8,86.4,75.8,71.4],['Toolathlon-Verified',70.4,68.5,47.2,27.8],['DeepPlanning',63.3,49.2,21.4,19.4]], reason:[['AIME 2026',95.2,94.5,94.7,89.2],['GPQA Diamond',89.5,89.2,83.5,91.3],['AA-LCR',84.0,82.0,83.3,72.0]], base:[['MMLU-Pro',66.9,null,null,59.3],['GPQA',62.0,null,null,41.5],['GSM8K',92.9,null,null,88.9],['Math',57.3,null,null,43.6],['CMMLU',86.8,null,null,76.9]] }; var cur='coding'; function drawBars(c){cur=c;var keys=c==='base'?['n','g']:['n','q','m','g']; var names=c==='base'?{n:'Nanbeige4.3-33B-Base',g:'Gemma4-31B-Base'}:{}; document.getElementById('nbgLegend').innerHTML=keys.map(function(k){return '<span><i style="background:'+M[k][1]+'"></i>'+(names[k]||M[k][0])+'</span>'}).join(''); var idx={n:1,q:2,m:3,g:4},h=''; D[c].forEach(function(r){h+='<div class="bench"><div class="bn">'+r[0]+'</div>'; keys.forEach(function(k){var v=r[idx[k]];h+='<div class="brow'+(k==='n'?' me':'')+'"><span class="lb">'+(names[k]||M[k][0])+'</span><div class="track"><div class="fill" data-w="'+v+'" style="background:'+M[k][1]+'"></div></div><span class="v">'+v.toFixed(1)+'</span></div>'}); h+='</div>'}); var box=document.getElementById('nbgBars');box.innerHTML=h; requestAnimationFrame(function(){requestAnimationFrame(function(){box.querySelectorAll('.fill').forEach(function(f){f.style.width=f.dataset.w+'%'})})}); resize();setTimeout(resize,300)} R.querySelectorAll('#nbgSeg button').forEach(function(b){b.onclick=function(){R.querySelectorAll('#nbgSeg button').forEach(function(x){x.classList.remove('on')});b.classList.add('on');drawBars(b.dataset.c)}}); drawBars('coding'); // calc var CT=[8192,16384,32768,65536,131072,262144,524288],wb=2,kb=2; function fmt(n){return n.toLocaleString('en-US')} function calc(){var ctx=CT[+document.getElementById('nbgCtx').value],seq=+document.getElementById('nbgSeq').value; document.getElementById('nbgCtxL').textContent=fmt(ctx)+' tokens';document.getElementById('nbgSeqL').textContent=seq; var w=33e9*wb/1e9,kv=52*8*128*2*kb*ctx*seq/1e9,t=w+kv; document.getElementById('nbgTot').textContent=t.toFixed(1)+' GB'; document.getElementById('nbgWv').textContent=w.toFixed(1)+' GB';document.getElementById('nbgKv').textContent=kv.toFixed(1)+' GB'; document.getElementById('nbgGpu').textContent=Math.ceil(t/80)+' × 80 GB'; document.getElementById('nbgBw').style.width=(w/t*100)+'%';document.getElementById('nbgBk').style.width=(kv/t*100)+'%'} ['nbgCtx','nbgSeq'].forEach(function(id){document.getElementById(id).addEventListener('input',calc)}); R.querySelectorAll('#nbgW button').forEach(function(b){b.onclick=function(){R.querySelectorAll('#nbgW button').forEach(function(x){x.classList.remove('on')});b.classList.add('on');wb=+b.dataset.w;calc()}}); R.querySelectorAll('#nbgK button').forEach(function(b){b.onclick=function(){R.querySelectorAll('#nbgK button').forEach(function(x){x.classList.remove('on')});b.classList.add('on');kb=+b.dataset.k;calc()}}); calc(); window.addEventListener('load',resize);window.addEventListener('resize',resize);setTimeout(resize,200); })(); " style="width:100%!important;height:600px;border:0!important;background:transparent!important;display:block;" loading="lazy"> window.addEventListener('message',function(e){if(e.data&&e.data.type==='mtp-nbg43-resize'){var f=document.getElementById('mtp-nbg43-frame');if(f&&e.data.height>100){f.style.height=e.data.height+'px';}}}); How does it compare with competitors’ own model cards? *Self-reported by Nanbeige. Competitor values come from each lab’s own Hugging Face model card. Some competitor figures in Nanbeige’s table differ from what those labs publish: Meta reports Muse-Glimmer-30B at 51.2 on SWE-Bench Pro and Gemma4-31B at 36.9. Nanbeige lists the reverse. Google lists Gemma 4 31B at 84.3 on GPQA Diamond. Nanbeige’s table shows 91.3. Meta reports Muse-Glimmer-30B at 51.7 on Terminal-Bench 2.1. Nanbeige measured 61.8. Different harnesses explain some gaps. The SWE-Bench Pro pair looks swapped. Treat these rows as provisional until independent runs appear. How do you deploy Nanbeige4.3-33B? The model card provides Transformers, SGLang and vLLM recipes. Its server examples use 4-way tensor parallelism with Qwen3 reasoning and tool-call parsers. Recommended agent settings are temperature 1.0, top_p 0.95 and 65,536 max new tokens. Reasoning and chat use temperature 0.6 and 262,144 max new tokens. By arithmetic from config.json, BF16 weights need about 66 GB. The KV cache adds about 208 KiB per token in BF16. A full 524,288-token sequence therefore needs roughly 112 GB of KV cache. At publication, no Hugging Face Inference Provider hosts the model. Key Takeaways 33B dense, Apache 2.0, 512K context, text-only. Self-reported 80.9 on Terminal-Bench 2.1 and 65.5 on DeepSWE. Leads 17 of 20 rows in Nanbeige’s own table. Several competitor figures conflict with official cards. Built as the teacher for the upcoming Nanbeige4.3-3B. Check out the model weights, the Base model and the launch post on X. 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 Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model appeared first on MarkTechPost.
Nanbeige LLM Lab has released Nanbeige4.3-33B on Hugging Face under the Apache 2.0 license. Nanbeige4.3-33B is a 33B dense language model built for agentic coding and office-style ‘cowork’ tasks. The lab sits inside the recruitment platform BOSS Zhipin. It also published Nanbeige4.3-33B-Base for custom post-training. The model supports a 524,288-token context window. In the lab’s own table, it leads 17 of 20 benchmarks against 3 open models near 30B.
TL;DR
- Size: 33B parameters, dense (no MoE), 524,288-token (512K) context, BF16 weights.
- Runs on: Multi-GPU servers. The card’s SGLang and vLLM recipes use 4-way tensor parallelism. BF16 weights need about 66 GB (our arithmetic). 1 community quantization is listed.
- Performance: Leads 17 of 20 rows in Nanbeige’s own comparison against Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B.
- Best: 80.9 on Terminal-Bench 2.1, versus 73.0 for Qwen3.8-27B.
- Worst: 23.8 on Terminal-Bench 4.0 (its lowest score). It also trails Qwen3.8-27B on SWE-Bench Pro, 60.3 vs 61.7.
What is Nanbeige4.3-33B?
Nanbeige4.3-33B is the flagship ‘teacher’ model of the Nanbeige4.3 series. The lab’s stated strategy is to build a strong teacher first. It then plans to distill agentic skills into smaller models, starting with Nanbeige4.3-3B.
How is Nanbeige4.3-33B built?
The config.json declares a standard LlamaForCausalLM architecture. There is no expert routing, so all 33B parameters run on every token.
- 52 layers, hidden size 7,168
- 56 query heads and 8 key-value heads (grouped-query attention), head dimension 128
- FFN intermediate size 22,016, vocabulary of 166,912 tokens
- RoPE theta of 120,000,000 and 524,288 max positions
The chat template exposes 3 controls. enable_thinking toggles reasoning and is on by default. preserve_thinking keeps reasoning from earlier turns. Nanbeige recommends it for multi-turn tool use and code agents. Passing tools enables tool calling, and an XML call format is recommended.
How does Nanbeige4.3-33B perform on benchmarks?
Nanbeige compared the model with Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B. Every model ran at its highest thinking effort, with shared hyperparameters and harnesses. All scores in this section are self-reported.
- Agentic coding: It scores 80.9 on Terminal-Bench 2.1, against 73.0 for Qwen3.8-27B. DeepSWE shows the widest gap: 65.5 versus 42.5. On Terminal-Bench 4.0, it scores 23.8 while Gemma4-31B scores 0.0.
- Agentic cowork: It posts 54.2 on Automation Bench versus 38.3, and 44.8 on JobBench versus 34.5. On YC-bench, simulated earnings reach $1,025,437 versus $641,290.
- Reasoning: It reports 95.2 on AIME 2026 and 84.0 on AA-LCR.
It loses 3 of 20 rows. Qwen3.8-27B leads SWE-Bench Pro (61.7 vs 60.3) and GDPval (86.4 vs 84.8). Gemma4-31B leads GPQA Diamond in Nanbeige’s table, 91.3 vs 89.5.
The Base model beats Gemma4-31B-Base on all 8 reported tasks. Examples include MMLU-Pro (66.9 vs 59.3) and GPQA (62.0 vs 41.5).
<iframe id="mtp-nbg43-frame" title="Nanbeige4.3-33B interactive explainer" srcdoc="<!doctype html>
<html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
<title>Nanbeige4.3-33B Explainer</title>
<style>
html,body{margin:0;padding:0;background:transparent}
#mtp-nbg43{–bg:#0b1d1c;–card:#112b29;–line:#24514b;–mint:#5fd3bd;–mint2:#8ee6d2;–ink:#e8f6f2;–mute:#9cbcb5;–qwen:#7c8cf8;–muse:#f0a35e;–gemma:#6aa9f4;–mtp:#76B900;
font-family:Inter,-apple-system,Segoe UI,Roboto,sans-serif!important;background:var(–bg)!important;color:var(–ink)!important;border:1px solid var(–line)!important;border-radius:16px;max-width:880px;margin:0 auto;padding:22px 24px 14px;box-sizing:border-box;overflow:hidden}
#mtp-nbg43 *{box-sizing:border-box}
#mtp-nbg43 hr,#mtp-nbg43 p:empty,#mtp-nbg43 del,#mtp-nbg43 s{display:none!important}
#mtp-nbg43 .hd{display:flex;align-items:center;justify-content:space-between;gap:12px;flex-wrap:wrap}
#mtp-nbg43 .tag{font-size:11px;letter-spacing:.12em;text-transform:uppercase;color:var(–mint)!important;font-weight:700}
#mtp-nbg43 h3{margin:4px 0 2px;font-size:22px;color:var(–ink)!important;font-weight:800;line-height:1.2}
#mtp-nbg43 .sub{color:var(–mute)!important;font-size:13px;margin:0}
#mtp-nbg43 .chips{display:flex;gap:6px;flex-wrap:wrap}
#mtp-nbg43 .chip{font-size:11px;padding:4px 9px;border-radius:999px;border:1px solid var(–line)!important;color:var(–mint2)!important;background:#0e2523!important;font-weight:600}
#mtp-nbg43 .tabs{display:flex;gap:6px;margin:18px 0 14px;border-bottom:1px solid var(–line)!important;flex-wrap:wrap}
#mtp-nbg43 .tab{background:transparent!important;border:0!important;color:var(–mute)!important;font:600 13px Inter,sans-serif;padding:9px 12px;cursor:pointer;border-bottom:2px solid transparent!important;margin-bottom:-1px;transition:color .2s}
#mtp-nbg43 .tab.on{color:var(–ink)!important;border-bottom-color:var(–mint)!important}
#mtp-nbg43 .pane{display:none;animation:nbgIn .35s ease}
#mtp-nbg43 .pane.on{display:block}
@keyframes nbgIn{from{opacity:0;transform:translateY(6px)}to{opacity:1;transform:none}}
/* flow */
#mtp-nbg43 .flow{display:grid;grid-template-columns:repeat(4,1fr);gap:10px;position:relative}
#mtp-nbg43 .node{background:var(–card)!important;border:1px solid var(–line)!important;border-radius:12px;padding:12px;min-height:96px;transition:all .35s;opacity:.45;cursor:pointer}
#mtp-nbg43 .node.on{opacity:1;border-color:var(–mint)!important;box-shadow:0 0 0 1px var(–mint),0 0 22px rgba(95,211,189,.25)}
#mtp-nbg43 .node .n{font-size:11px;color:var(–mint)!important;font-weight:700}
#mtp-nbg43 .node .t{font-weight:700;font-size:14px;margin:4px 0;color:var(–ink)!important}
#mtp-nbg43 .node .d{font-size:12px;color:var(–mute)!important;line-height:1.4}
#mtp-nbg43 .rail{height:4px!important;background:var(–line)!important;border-radius:4px;margin:12px 0 0;overflow:hidden}
#mtp-nbg43 .rail i{display:block;height:100%;width:25%;background:linear-gradient(90deg,var(–mint),var(–mint2));transition:width .5s ease}
#mtp-nbg43 .detail{background:#0e2523!important;border:1px solid var(–line)!important;border-radius:12px;padding:14px 16px;margin-top:12px;font-size:13.5px;line-height:1.55;min-height:92px}
#mtp-nbg43 .detail b{color:var(–mint2)!important}
#mtp-nbg43 .ctl{display:flex;gap:8px;margin-top:12px;align-items:center}
#mtp-nbg43 .btn{background:var(–mint)!important;color:#06201c!important;border:0!important;border-radius:8px;font:700 13px Inter,sans-serif;padding:8px 14px;cursor:pointer}
#mtp-nbg43 .btn.ghost{background:transparent!important;color:var(–mint2)!important;border:1px solid var(–line)!important}
#mtp-nbg43 .spec{display:grid;grid-template-columns:repeat(4,1fr);gap:8px;margin-top:14px}
#mtp-nbg43 .spec div{background:var(–card)!important;border:1px solid var(–line)!important;border-radius:10px;padding:10px}
#mtp-nbg43 .spec b{display:block;font-size:18px;color:var(–mint2)!important}
#mtp-nbg43 .spec span{font-size:11px;color:var(–mute)!important}
/* bars */
#mtp-nbg43 .seg{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:12px}
#mtp-nbg43 .seg button{background:#0e2523!important;border:1px solid var(–line)!important;color:var(–mute)!important;font:600 12px Inter,sans-serif;padding:6px 11px;border-radius:999px;cursor:pointer}
#mtp-nbg43 .seg button.on{background:var(–mint)!important;color:#06201c!important;border-color:var(–mint)!important}
#mtp-nbg43 .legend{display:flex;gap:14px;flex-wrap:wrap;font-size:12px;color:var(–mute)!important;margin-bottom:10px}
#mtp-nbg43 .legend i{display:inline-block;width:10px;height:10px;border-radius:3px;margin-right:5px;vertical-align:-1px}
#mtp-nbg43 .bench{margin-bottom:12px}
#mtp-nbg43 .bench .bn{font-size:13px;font-weight:700;margin-bottom:5px;color:var(–ink)!important}
#mtp-nbg43 .brow{display:grid;grid-template-columns:110px 1fr 54px;gap:8px;align-items:center;font-size:11.5px;margin:2px 0}
#mtp-nbg43 .brow .lb{color:var(–mute)!important;white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
#mtp-nbg43 .track{height:12px!important;background:#0e2523!important;border-radius:6px;overflow:hidden}
#mtp-nbg43 .fill{height:100%;width:0;border-radius:6px;transition:width .9s cubic-bezier(.2,.8,.2,1)}
#mtp-nbg43 .brow .v{text-align:right;font-variant-numeric:tabular-nums;color:var(–ink)!important}
#mtp-nbg43 .brow.me .v,#mtp-nbg43 .brow.me .lb{color:var(–mint2)!important;font-weight:700}
#mtp-nbg43 .note{font-size:11.5px;color:var(–mute)!important;margin-top:6px;line-height:1.45}
/* calc */
#mtp-nbg43 .calc{display:grid;grid-template-columns:1.1fr 1fr;gap:16px}
#mtp-nbg43 .field{margin-bottom:14px}
#mtp-nbg43 .field label{display:flex;justify-content:space-between;font-size:12.5px;color:var(–mute)!important;margin-bottom:6px}
#mtp-nbg43 .field label b{color:var(–mint2)!important}
#mtp-nbg43 input[type=range]{width:100%;accent-color:#5fd3bd}
#mtp-nbg43 .stack{background:var(–card)!important;border:1px solid var(–line)!important;border-radius:12px;padding:14px}
#mtp-nbg43 .stackbar{display:flex;height:22px!important;border-radius:6px;overflow:hidden;background:#0e2523!important;margin:8px 0 10px}
#mtp-nbg43 .stackbar i{display:block;height:100%;transition:width .5s ease}
#mtp-nbg43 .total{font-size:30px;font-weight:800;color:var(–ink)!important}
#mtp-nbg43 .kv{display:flex;justify-content:space-between;font-size:12.5px;margin:4px 0;color:var(–mute)!important}
#mtp-nbg43 .kv b{color:var(–ink)!important}
/* caveats */
#mtp-nbg43 table{width:100%;border-collapse:collapse;font-size:12.5px}
#mtp-nbg43 th,#mtp-nbg43 td{border-bottom:1px solid var(–line)!important;padding:8px 6px;text-align:left;color:var(–ink)!important;background:transparent!important}
#mtp-nbg43 th{color:var(–mint)!important;font-size:11px;text-transform:uppercase;letter-spacing:.06em}
#mtp-nbg43 td.bad{color:#f6b26b!important;font-weight:700}
#mtp-nbg43 .flag{display:inline-block;font-size:10.5px;padding:2px 7px;border-radius:999px;background:#3a2a12!important;color:#f6b26b!important;font-weight:700}
#mtp-nbg43 .ft{display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:8px;border-top:1px solid var(–line)!important;margin-top:16px;padding-top:10px;font-size:11px;color:var(–mute)!important}
#mtp-nbg43 .ft a{color:var(–mute)!important;text-decoration:underline}
#mtp-nbg43 .brand{color:#76B900!important;font-weight:700}
#mtp-nbg43 pre,#mtp-nbg43 code{background:#0e2523!important;color:var(–mint2)!important;border:0!important;font-family:ui-monospace,Menlo,monospace;font-size:12px;padding:1px 5px;border-radius:4px}
@media (max-width:640px){
#mtp-nbg43{padding:16px 14px 10px}
#mtp-nbg43 h3{font-size:18px}
#mtp-nbg43 .flow,#mtp-nbg43 .spec{grid-template-columns:1fr 1fr}
#mtp-nbg43 .calc{grid-template-columns:1fr}
#mtp-nbg43 .brow{grid-template-columns:86px 1fr 48px}
#mtp-nbg43 .tab{padding:8px 8px;font-size:12px}
}
</style></head><body>
<div id="mtp-nbg43">
<div class="hd">
<div>
<div class="tag">Nanbeige LLM Lab · Open weights</div>
<h3>Nanbeige4.3-33B, explained</h3>
<p class="sub">A 33B dense teacher model for agentic coding and cowork tasks</p>
</div>
<div class="chips"><span class="chip">33B dense</span><span class="chip">512K context</span><span class="chip">Apache 2.0</span><span class="chip">Text-only</span></div>
</div>
<div class="tabs" role="tablist">
<button class="tab on" data-p="p1">How it works</button>
<button class="tab" data-p="p2">Benchmarks</button>
<button class="tab" data-p="p3">Memory math</button>
<button class="tab" data-p="p4">Caveats</button>
</div>
<div class="pane on" id="p1">
<div class="flow">
<div class="node on" data-s="0"><div class="n">STEP 1</div><div class="t">Base model</div><div class="d">Nanbeige4.3-33B-Base, released for custom post-training</div></div>
<div class="node" data-s="1"><div class="n">STEP 2</div><div class="t">Teacher agent</div><div class="d">Post-trained Nanbeige4.3-33B with thinking and tools</div></div>
<div class="node" data-s="2"><div class="n">STEP 3</div><div class="t">Agent harness</div><div class="d">Runs inside harnesses like Terminus 2, mini-swe-agent, Claude Code</div></div>
<div class="node" data-s="3"><div class="n">STEP 4</div><div class="t">Distillation</div><div class="d">Planned: agentic skills move into Nanbeige4.3-3B</div></div>
</div>
<div class="rail"><i id="nbgRail"></i></div>
<div class="detail" id="nbgDetail"></div>
<div class="ctl"><button class="btn ghost" id="nbgPrev"> Back</button><button class="btn" id="nbgNext">Next step </button><button class="btn ghost" id="nbgPlay">Auto-play</button></div>
<div class="spec">
<div><b>52</b><span>layers</span></div>
<div><b>56 / 8</b><span>query / KV heads (GQA)</span></div>
<div><b>7,168</b><span>hidden size</span></div>
<div><b>524,288</b><span>max positions</span></div>
</div>
</div>
<div class="pane" id="p2">
<div class="seg" id="nbgSeg">
<button class="on" data-c="coding">Agentic coding</button>
<button data-c="cowork">Agentic cowork</button>
<button data-c="reason">Reasoning</button>
<button data-c="base">Base model</button>
</div>
<div class="legend" id="nbgLegend"></div>
<div id="nbgBars"></div>
<p class="note">All scores self-reported by Nanbeige on its Hugging Face model cards, at each model’s highest thinking effort with shared harnesses. Not independently verified.</p>
</div>
<div class="pane" id="p3">
<div class="calc">
<div>
<div class="field"><label>Context length <b id="nbgCtxL">131,072 tokens</b></label><input type="range" id="nbgCtx" min="0" max="6" step="1" value="4"></div>
<div class="field"><label>Concurrent sequences <b id="nbgSeqL">1</b></label><input type="range" id="nbgSeq" min="1" max="8" step="1" value="1"></div>
<div class="seg" id="nbgW"><button class="on" data-w="2">Weights BF16</button><button data-w="1">Weights 8-bit</button></div>
<div class="seg" id="nbgK"><button class="on" data-k="2">KV cache BF16</button><button data-k="1">KV cache FP8</button></div>
<p class="note">Arithmetic from config.json: 52 layers × 8 KV heads × 128 dims × 2 (K and V) × bytes per value. That is 212,992 bytes (about 208 KiB) per token in BF16. Weights use 33B parameters. Excludes activations and framework overhead.</p>
</div>
<div class="stack">
<div class="kv"><span>Estimated GPU memory</span></div>
<div class="total" id="nbgTot">0 GB</div>
<div class="stackbar"><i id="nbgBw" style="background:#5fd3bd"></i><i id="nbgBk" style="background:#7c8cf8"></i></div>
<div class="kv"><span><i style="display:inline-block;width:9px;height:9px;background:#5fd3bd;border-radius:2px"></i> Weights</span><b id="nbgWv"></b></div>
<div class="kv"><span><i style="display:inline-block;width:9px;height:9px;background:#7c8cf8;border-radius:2px"></i> KV cache</span><b id="nbgKv"></b></div>
<div class="kv"><span>Per 80 GB GPU (ideal fit)</span><b id="nbgGpu"></b></div>
<p class="note">The model card’s SGLang and vLLM recipes use 4-way tensor parallelism.</p>
</div>
</div>
</div>
<div class="pane" id="p4">
<table>
<tr><th>Row in Nanbeige table</th><th>Nanbeige lists</th><th>Official card</th><th></th></tr>
<tr><td>Muse-Glimmer-30B, SWE-Bench Pro</td><td class="bad">36.9</td><td>51.2 (Meta)</td><td><span class="flag">looks swapped</span></td></tr>
<tr><td>Gemma4-31B, SWE-Bench Pro</td><td class="bad">51.2</td><td>36.9 (Meta-reported)</td><td><span class="flag">looks swapped</span></td></tr>
<tr><td>Gemma4-31B, GPQA Diamond</td><td class="bad">91.3</td><td>84.3 (Google)</td><td><span class="flag">conflict</span></td></tr>
<tr><td>Muse-Glimmer-30B, Terminal-Bench 2.1</td><td class="bad">61.8</td><td>51.7 (Meta)</td><td><span class="flag">harness?</span></td></tr>
<tr><td>Qwen3.8-27B, GPQA Diamond</td><td>89.2</td><td>89.2 (Qwen)</td><td>match</td></tr>
<tr><td>Muse-Glimmer-30B, AIME 2026</td><td>94.7</td><td>94.7 (Meta)</td><td>match</td></tr>
</table>
<p class="note">Nanbeige re-ran every model with its own harness, which explains some gaps. Other limits: text-only input, dense compute on every token, and Nanbeige4.3-3B is not yet released.</p>
</div>
<div class="ft"><span>Sources: <a href=" target="_blank" rel="noopener">Nanbeige model card</a> · <a href=" target="_blank" rel="noopener">Base card</a> · competitor HF cards</span><span class="brand">© Marktechpost</span></div>
</div>
<script>
(function(){
var R=document.getElementById(‘mtp-nbg43′);
function resize(){try{parent.postMessage({type:’mtp-nbg43-resize’,height:R.offsetHeight+40},’*’)}catch(e){}}
// tabs
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.remove(‘on’)});
document.getElementById(t.dataset.p).classList.add(‘on’);
if(t.dataset.p===’p2′)drawBars(cur);
setTimeout(resize,50);setTimeout(resize,400);
})});
// flow
var steps=[
‘<b>Base model.</b> Nanbeige4.3-33B-Base is a dense LlamaForCausalLM checkpoint. Nanbeige reports it beats Gemma4-31B-Base on all 8 listed tasks, for example MMLU-Pro 66.9 vs 59.3.’,
‘<b>Teacher agent.</b> The post-trained model thinks by default (<code>enable_thinking</code>). <code>preserve_thinking</code> keeps earlier reasoning for multi-turn tool use. Nanbeige recommends the XML tool-call format.’,
‘<b>Agent harness.</b> Nanbeige evaluated it in Terminus 2, mini-swe-agent v2.4.6 and the Claude Code harness. Recommended agent sampling: temperature 1.0, top_p 0.95, 65,536 max new tokens.’,
‘<b>Distillation.</b> The 33B model is the teacher for the Nanbeige4.3 series. The first planned student is Nanbeige4.3-3B. No release date is given.’
];
var s=0,timer=null;
function setStep(i){s=(i+4)%4;R.querySelectorAll(‘.node’).forEach(function(n,k){n.classList.toggle(‘on’,k<=s)});
document.getElementById(‘nbgRail’).style.width=((s+1)*25)+’%’;document.getElementById(‘nbgDetail’).innerHTML=steps[s];resize()}
R.querySelectorAll(‘.node’).forEach(function(n){n.addEventListener(‘click’,function(){stop();setStep(+n.dataset.s)})});
function stop(){if(timer){clearInterval(timer);timer=null;document.getElementById(‘nbgPlay’).textContent=’Auto-play’}}
document.getElementById(‘nbgNext’).onclick=function(){stop();setStep(s+1)};
document.getElementById(‘nbgPrev’).onclick=function(){stop();setStep(s-1)};
document.getElementById(‘nbgPlay’).onclick=function(){if(timer){stop();return}this.textContent=’Pause’;timer=setInterval(function(){setStep(s+1)},2200)};
setStep(0);
// bars
var M={n:[‘Nanbeige4.3-33B’,’#5fd3bd’],q:[‘Qwen3.8-27B’,’#7c8cf8′],m:[‘Muse-Glimmer-30B’,’#f0a35e’],g:[‘Gemma4-31B’,’#6aa9f4′]};
var D={
coding:[[‘Terminal-Bench 2.1’,80.9,73.0,61.8,47.2],[‘DeepSWE’,65.5,42.5,4.4,2.7],[‘SWE-Bench Pro’,60.3,61.7,36.9,51.2],[‘SkillsBench’,61.1,56.4,39.7,28.9],[‘Terminal-Bench 4.0’,23.8,6.3,1.6,0.0]],
cowork:[[‘Automation Bench’,54.2,38.3,15.2,9.7],[‘JobBench’,44.8,34.5,16.6,9.0],[‘GDPval’,84.8,86.4,75.8,71.4],[‘Toolathlon-Verified’,70.4,68.5,47.2,27.8],[‘DeepPlanning’,63.3,49.2,21.4,19.4]],
reason:[[‘AIME 2026’,95.2,94.5,94.7,89.2],[‘GPQA Diamond’,89.5,89.2,83.5,91.3],[‘AA-LCR’,84.0,82.0,83.3,72.0]],
base:[[‘MMLU-Pro’,66.9,null,null,59.3],[‘GPQA’,62.0,null,null,41.5],[‘GSM8K’,92.9,null,null,88.9],[‘Math’,57.3,null,null,43.6],[‘CMMLU’,86.8,null,null,76.9]]
};
var cur=’coding’;
function drawBars(c){cur=c;var keys=c===’base’?[‘n’,’g’]:[‘n’,’q’,’m’,’g’];
var names=c===’base’?{n:’Nanbeige4.3-33B-Base’,g:’Gemma4-31B-Base’}:{};
document.getElementById(‘nbgLegend’).innerHTML=keys.map(function(k){return ‘<span><i style="background:’+M[k][1]+’"></i>’+(names[k]||M[k][0])+'</span>’}).join(”);
var idx={n:1,q:2,m:3,g:4},h=”;
D[c].forEach(function(r){h+='<div class="bench"><div class="bn">’+r[0]+'</div>’;
keys.forEach(function(k){var v=r[idx[k]];h+='<div class="brow’+(k===’n’?’ me’:”)+’"><span class="lb">’+(names[k]||M[k][0])+'</span><div class="track"><div class="fill" data-w="’+v+’" style="background:’+M[k][1]+’"></div></div><span class="v">’+v.toFixed(1)+'</span></div>’});
h+='</div>’});
var box=document.getElementById(‘nbgBars’);box.innerHTML=h;
requestAnimationFrame(function(){requestAnimationFrame(function(){box.querySelectorAll(‘.fill’).forEach(function(f){f.style.width=f.dataset.w+’%’})})});
resize();setTimeout(resize,300)}
R.querySelectorAll(‘#nbgSeg button’).forEach(function(b){b.onclick=function(){R.querySelectorAll(‘#nbgSeg button’).forEach(function(x){x.classList.remove(‘on’)});b.classList.add(‘on’);drawBars(b.dataset.c)}});
drawBars(‘coding’);
// calc
var CT=[8192,16384,32768,65536,131072,262144,524288],wb=2,kb=2;
function fmt(n){return n.toLocaleString(‘en-US’)}
function calc(){var ctx=CT[+document.getElementById(‘nbgCtx’).value],seq=+document.getElementById(‘nbgSeq’).value;
document.getElementById(‘nbgCtxL’).textContent=fmt(ctx)+’ tokens’;document.getElementById(‘nbgSeqL’).textContent=seq;
var w=33e9*wb/1e9,kv=52*8*128*2*kb*ctx*seq/1e9,t=w+kv;
document.getElementById(‘nbgTot’).textContent=t.toFixed(1)+’ GB’;
document.getElementById(‘nbgWv’).textContent=w.toFixed(1)+’ GB’;document.getElementById(‘nbgKv’).textContent=kv.toFixed(1)+’ GB’;
document.getElementById(‘nbgGpu’).textContent=Math.ceil(t/80)+’ × 80 GB’;
document.getElementById(‘nbgBw’).style.width=(w/t*100)+’%’;document.getElementById(‘nbgBk’).style.width=(kv/t*100)+’%’}
[‘nbgCtx’,’nbgSeq’].forEach(function(id){document.getElementById(id).addEventListener(‘input’,calc)});
R.querySelectorAll(‘#nbgW button’).forEach(function(b){b.onclick=function(){R.querySelectorAll(‘#nbgW button’).forEach(function(x){x.classList.remove(‘on’)});b.classList.add(‘on’);wb=+b.dataset.w;calc()}});
R.querySelectorAll(‘#nbgK button’).forEach(function(b){b.onclick=function(){R.querySelectorAll(‘#nbgK button’).forEach(function(x){x.classList.remove(‘on’)});b.classList.add(‘on’);kb=+b.dataset.k;calc()}});
calc();
window.addEventListener(‘load’,resize);window.addEventListener(‘resize’,resize);setTimeout(resize,200);
})();
</script>
</body></html>
” style=”width:100%!important;height:600px;border:0!important;background:transparent!important;display:block;” loading=”lazy”>
window.addEventListener(‘message’,function(e){if(e.data&&e.data.type===’mtp-nbg43-resize’){var f=document.getElementById(‘mtp-nbg43-frame’);if(f&&e.data.height>100){f.style.height=e.data.height+’px’;}}});
How does it compare with competitors’ own model cards?
*Self-reported by Nanbeige. Competitor values come from each lab’s own Hugging Face model card.
Some competitor figures in Nanbeige’s table differ from what those labs publish:
- Meta reports Muse-Glimmer-30B at 51.2 on SWE-Bench Pro and Gemma4-31B at 36.9. Nanbeige lists the reverse.
- Google lists Gemma 4 31B at 84.3 on GPQA Diamond. Nanbeige’s table shows 91.3.
- Meta reports Muse-Glimmer-30B at 51.7 on Terminal-Bench 2.1. Nanbeige measured 61.8.
Different harnesses explain some gaps. The SWE-Bench Pro pair looks swapped. Treat these rows as provisional until independent runs appear.
How do you deploy Nanbeige4.3-33B?
The model card provides Transformers, SGLang and vLLM recipes. Its server examples use 4-way tensor parallelism with Qwen3 reasoning and tool-call parsers. Recommended agent settings are temperature 1.0, top_p 0.95 and 65,536 max new tokens. Reasoning and chat use temperature 0.6 and 262,144 max new tokens.
By arithmetic from config.json, BF16 weights need about 66 GB. The KV cache adds about 208 KiB per token in BF16. A full 524,288-token sequence therefore needs roughly 112 GB of KV cache. At publication, no Hugging Face Inference Provider hosts the model.
Key Takeaways
- 33B dense, Apache 2.0, 512K context, text-only.
- Self-reported 80.9 on Terminal-Bench 2.1 and 65.5 on DeepSWE.
- Leads 17 of 20 rows in Nanbeige’s own table.
- Several competitor figures conflict with official cards.
- Built as the teacher for the upcoming Nanbeige4.3-3B.
Check out the model weights, the Base model and the launch post on X. 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 Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model appeared first on MarkTechPost.
Source: MarkTechPost
