StepFun has released Step 5 Preview as open weights. FP8 and BF16 checkpoints are live on Hugging Face and ModelScope. Step 5 Preview is a sparse Mixture-of-Experts model from StepFun built for long-horizon agentic work. It has 600B total parameters, 27B active per token and up to 1M tokens of context. For developers, this is a new self-hostable option for coding agents, document analysis and finance workflows.
TL;DR
- Size: 600B total, 27B active per token, up to 1M-token context.
- Runs on: data-center GPUs only. FP8 needs 4x B300 or 8x H200. BF16 needs 8x B300 or 16x H200.
- Performance: near open-weight peers GLM-5.3 and Kimi K3 on coding. Behind GPT-6 Astra and Claude Opus 5.
- Best: 88.3% on AA-LCR long-context reasoning (Artificial Analysis).
- Worst: 33.3% on Terminal-Bench 4.0, versus 41.9% for GLM-5.3.
What is Step 5 Preview?
Step 5 Preview is StepFun’s flagship model for agentic work. StepFun positions it for software engineering, professional knowledge work and finance. The model navigates large codebases, coordinates multi-file changes, runs code and iterates on results. It also analyzes long documents, charts, screenshots and video to produce structured reports.
The API launched September 20. Weights followed on October 14.
How does the architecture work?
Only 27B of the 600B parameters activate per token. StepFun’s tech team adds the internals:
- Depth: 92 transformer blocks in a narrow, deep design.
- Attention mix: gated sliding-window attention and full attention layers, interleaved 3:1.
- Sparse GQA: full-attention layers pick 4,096 history tokens per query, scored in blocks of 8.
- Speculative decoding: an MTP-3 head gives roughly 30% end-to-end speedup.
- Training: FP8 mixed precision, with RL on million-level verifiable agentic environments.
How does Step 5 Preview perform on benchmarks?
StepFun ran its own evaluations at the High reasoning setting. Rivals in its chart ran at Max effort, so the comparison is not like-for-like.
Results reported on the Hugging Face card:
- DeepSWE v1.1: 67.7 (SWE-agent harness, temperature 1.0, top_p 0.95)
- Terminal-Bench 4.0: 33.3
- Agents’ Last Exam (ALE-CLI): 29.5
Independent numbers come from Artificial Analysis via OpenRouter. The model scores 43.7 on the Intelligence Index. It hits 88.3% on AA-LCR and 46.5% on Humanity’s Last Exam. GDPval-AA lands at 54.3%.
Vals AI results show clear weak spots. SRE Bench is 0.4%. Harvey’s Legal Agent Benchmark is 9.2%.
On finance, StepFun’s chart shows 66.4% on FrontierFinance. That beats GLM-5.3 and Kimi K3, but trails Claude Opus 5 at 69.7%.
‘});
document.getElementById(‘s5bars’).innerHTML=h;document.getElementById(‘s5bn’).textContent=B[k].n;
setTimeout(function(){R.querySelectorAll(‘.fil’).forEach(function(f){f.style.width=f.dataset.w+’%’})},40);post()}
R.querySelectorAll(‘[data-b]’).forEach(function(b){b.onclick=function(){bars(b.dataset.b)}});
var H=[[‘FP8 · B300′,’vLLM or SGLang, 1 node’,4],[‘FP8 · H200′,’vLLM or SGLang, 1 node’,8],[‘BF16 · B300′,’vLLM, 1 node’,8],[‘BF16 · H200′,’vLLM, 2 nodes’,16]];
var hw=document.getElementById(‘s5hw’),hd=document.getElementById(‘s5hwd’);
H.forEach(function(x,i){var c=document.createElement(‘div’);c.className=”hwc”;c.innerHTML=’
‘+x[0]+’
‘+x[1]+’, ‘+x[2]+’ GPUs
‘;
c.onclick=function(){hw.querySelectorAll(‘.hwc’).forEach(function(y){y.classList.remove(‘on’);y.querySelector(‘.gpus’).innerHTML=”});c.classList.add(‘on’);
var g=c.querySelector(‘.gpus’);for(var k=0;k
2-node setups add –nnodes 2 and node-rank flags. See the model card.
‘:”);post()};hw.appendChild(c)});
hw.children[0].click();
window.addEventListener(‘load’,post);setTimeout(post,300);
})();