StepFun joins the frontier: a previously non-frontier Chinese lab (StepFun) released a Kimi K3-level model, 3 times cheaper per Artificial Analysis
StepFun, a Chinese lab, has released its Step 5 Preview model, which Artificial Analysis indicates performs at a Kimi K3-level. This new model is reportedly three times cheaper than Kimi K3, positioning StepFun firmly on the Pareto line for cost-intelligence. This release suggests the competitive race in AI development continues, with new players emerging to challenge established models on both performance and price. Meanwhile, Google's global lab ranking reportedly dropped from #7 to #8.
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年9月19日 05:18 UTC
收录当时偏移：UTC+02026年9月19日 17:00 UTC
- 发布
- 2026年9月19日 05:18
- 收录
- 2026年9月19日 17:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
So the race, despite everything, continues. StepFun releases Step 5 Preview, scoring equally to Kimi K3 (max) (44), at 1/3 the price, and stands firmly on the Pareto line of cost/intelligence.
However, as a closed-source model, it remains to be seen how competitive this will be against other Chinese labs. Nonetheless, it seems that another Chinese lab has reached the frontier and shows that China has not stopped progressing in the meantime. It remains to be seen if they'll stay there in the coming months.
On another note, Google fell from #7 lab in the world to #8 lol get rekt