{"ok":true,"trend":{"id":192226,"platform":"mastodon","region":"global","key":"chinese ai lab stepfun has unveiled step 5 preview, a sparse mixture of experts model with roughly 600 billion total par","title":"Chinese AI lab StepFun has unveiled Step 5 Preview, a sparse mixture of experts model with roughly 600 billion total par","url":"https://universeodon.com/@china/117346804681270631","first_seen":"2026-09-28T04:44:42.616580Z","last_seen":"2026-09-28T04:44:42.616580Z","last_rank":5,"peak_rank":5,"last_volume":1,"peak_volume":1,"seen_count":1,"score":66.39,"category_hint":"ai","section":"technology","category":"ai","summary":"Chinese AI lab StepFun has unveiled Step 5 Preview, a sparse mixture-of-experts model with roughly 600 billion total parameters and about 27 billion active per token. The model supports a one million token context window and is designed for long-horizon agent work, with open weights. Observers are noting the specifications and what the release means for competition among Chinese AI labs and open-weight model development.","why":"A major new open-weight AI model release from a leading Chinese lab draws attention amid fierce global AI competition","tone":"neutral","entities":["StepFun","Step 5 Preview","China"],"summarized_at":"2026-09-28T04:44:53.941332Z","meta":{"tag":"ai","via":"scan","kind":"status","lang":"en","generic":0.04,"instance":"mastodon.social","tag_uses":1269},"nw":0.72,"promo":0.03,"kind":"news_event","importance":1.99,"hidden":false,"hide_reason":null,"judged_at":"2026-09-28T04:44:43.344079Z","title_en":"Chinese AI lab StepFun unveils Step 5 Preview model","section_name":"Technology","category_name":"AI","timeline":[{"captured_at":"2026-09-28T04:44:42.616580Z","rank":5,"volume":1}],"posts":[{"platform":"mastodon","url":"https://universeodon.com/@china/117346804681270631","author":"china@universeodon.com","title":null,"snippet":"Chinese AI lab StepFun has unveiled Step 5 Preview, a sparse mixture of experts model with roughly 600 billion total parameters and about 27 billion active per token. The model supports a one million token context window and is designed for long-horizon agent work. Open weights…","posted_at":"2026-09-28T04:35:39Z","likes":1}],"elsewhere":[],"window":"7d"}}