StepFun launches Step 5 Preview: 600B MoE, 27B active, 1M context at 65% lower cost
realmrfakename · x · 2026-09-21
China's StepFun has launched Step 5 Preview, a 600B-parameter MoE model with just 27B active parameters and a 1M-token context window.
- Scores 44 on Artificial Analysis, matching Kimi K3 (max) while costing roughly 65% less per task, putting it on the cost/intelligence Pareto frontier
- Available via API now, with open weights promised for October 15
Related event: StepFun Launches Step 5 Preview: 600B MoE Flagship, Open-Weights Oct 15(16 posts)→
More from Models
- Tobi Lütke: local Dell server runs DeepSeek 4.1 Flash at ~300 tok/s, a billion tokens a month — BLUECOW009 · 2026-09-21
- LLMs still struggle at niche-domain labeling: a cluster-then-LLM pipeline with brittle second-pass labels — FanaHOVA · 2026-09-21
- A "decide" Primitive: Browser On-Device AI as a Native JS Control Flow — cocktailpeanut · 2026-09-21
- Rumor: GPT-6 Sol Lands Tuesday, Internal 'Bel' Deemed AGI; Opus 5.5 May Drop Monday — imjustnewatai · 2026-09-21
- Jev underperforms: researchers find better alternatives to GLiNER2-class extractors — airesearch12 · 2026-09-21
- Nym rebuilds its agent around Jev, a fast classifier model, for speed and cost gains — moyix · 2026-09-21