Step 5 Preview
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 9
- StepFun to join LA Hacks 2026 and host a Step 5 Preview workshop
StepFun announced it will be at UCLA on October 17–18 for LA Hacks 2026 to meet builders, exchange ideas and explore what's possible with Step 5 Preview. The company will also host a workshop on Step 5 Preview on Saturday, October 17, 8–9 PM PT at the James West Alumni Center.
阶跃星辰 StepFun · 🔥 8 - Step 5 Preview reaches #2 on OpenRouter weekly leaderboard one week after launch
StepFun says Step 5 Preview has climbed to second place on OpenRouter's weekly leaderboard one week after launch, with 19.5T tokens used in that period. The company adds that the model has held the #2 spot on the daily leaderboard since Oct 10 and thanks developers building with it; no further benchmark or capability details were given.
阶跃星辰 StepFun · 🔥 8 - Moore Threads completes Day-0 support for Step 5 Preview on MTT S5000
StepFun open-sourced its new Step 5 Preview model on October 15, and Moore Threads said the same day that its MTT S5000 compute card and accompanying MUSA software stack had already completed Day-0 adaptation for the model. According to the source, Step 5 Preview uses a sparse MoE architecture with 600B total parameters, 27B activated per token, and native support for contexts up to 1 million tokens. Moore Threads says the adaptation covers key areas including hybrid attention, mixed-precision inference and speculative decoding, giving developers a path to deploy long-context models on domestic compute.
AIbase AI新闻 · 🔥 6 - Step 5 Preview goes live on DeepInfra with 1M-token context
StepFun said its Step 5 Preview model is now available on DeepInfra, giving developers another route to call the preview release. According to the two companies, it is a 600B-parameter open-weight MoE with 27B active parameters, a 1M-token context window, text and vision input and FP8 precision, priced at $1.00 per 1M input tokens, $2.70 per 1M output tokens and $0.05 per 1M cached tokens. StepFun describes it as built for long-horizon agents in coding, frontend and finance work.
阶跃星辰 StepFun · 🔥 5 - StepFun opens Step 5 Preview weights: 600B MoE, 1M-token context
StepFun released BF16 and FP8 weights for Step 5 Preview, its flagship long-horizon agent model, on Hugging Face and ModelScope, plus a technical blog and deployment guides; the sparse MoE model has 600B total and 27B active parameters, supports up to a 1M-token context with text and image input, and targets software engineering, professional knowledge work and finance. According to StepFun, the design uses a 92-layer narrow-deep architecture with 3:1 gated sliding-window to sparse global attention and Sparse GQA, where 8-token block indexing cuts index and top-k selection cost to one eighth, plus FP32 key states, fine-grained gradient clipping and long-horizon RL that trains context compression; it can be served via vLLM or SGLang with an OpenAI-compatible API, or run offline with Transformers. StepFun-cited data puts it at 44 on Artificial Analysis's Intelligence Index and second on OpenRouter's weekly and daily token charts (about 18.8T tokens in seven days), at roughly $0.71 per task versus $5.86 for Claude Opus 5, and reports it above Kimi K3 and GLM-5.3 on ALE-CLI, FrontierFinance and DRACO, behind GPT-6 Astra or Claude Opus 5.
阶跃星辰 · Hugging Face · 🔥 41 - StepFun's Step 5 Preview lands on Tencent WorkBuddy with two weeks free
StepFun announced on October 12 that its flagship Step 5 Preview model is now available on Tencent's all-scenario AI office workspace WorkBuddy, with exclusive free access from October 12 to October 25, 2026. WorkBuddy users can use the model for deep research, information retrieval, data analysis, and document and PPT generation, including sustained execution of multi-step tasks. Step 5 Preview launched on September 20 with a sparse MoE architecture, 600B total parameters and 27B activated, a 1M-token context window and native text and vision input; StepFun says it plans to open-source the full weights on October 15.
IT之家 AI · 🔥 0 - STEPX launches Neo, billed as first large-model-native agent phone, from ¥3,999
Shanghai-based terminal brand STEPX unveiled the STEPX Neo Pioneer Edition on Oct 13, describing it as the first large-model-native agent phone: a Snapdragon 8-series device with 16GB+256GB and a bundled “model + software + hardware” price of ¥3,999 covering the handset, Step AOS 1.0, Step Studio and the Step model family, with pre-orders starting that day and first sales on Oct 16. The phone centers on the Amoo personal agent and the agent-native Step AOS 1.0, which adds dual-domain memory for the user and the agent, task scheduling and a “trusted, visible, controllable, reversible” safety framework, while a six-microphone array, Whisper 1.0 pickup algorithm, low-power audio chip and a MiniLED dot-matrix rear screen handle sensing and feedback, and Edge on-device plus Pro/Flash cloud models split work by task. STEPX says Amoo is connected to roughly 100 partners including Alipay, Tencent (WorkBuddy, WeCom, WeRead), Ctrip, Didi, CapCut, Amap and Meituan, and that Step 5 Preview is free to try inside WorkBuddy until Oct 25; buyers get 12 months of 10 billion tokens per month, and the first 10,000 qualifying users receive a five-year zero-cost upgrade ticket.
智东西 · 🔥 0 - Step 5 Preview hits OpenRouter's New & Trending leaderboard
StepFun said Step 5 Preview has landed on OpenRouter's New & Trending leaderboard. The company thanked users for trying it and partners for bringing it into their workflows. The source gave no benchmark figures or ranking details.
阶跃星辰 StepFun · 🔥 0 - Step 5 Preview free for a week across several coding tools
StepFun says its flagship model for agentic and professional work, Step 5 Preview, is free for one week in OpenCode, Nous Portal (with Hermes Agent), Cline and Kilo. Partners describe it as a 600B-parameter MoE with 27B active per token, a 1M-token context and vision, with OpenCode noting zero data retention. Cline claims it scores ahead of Kimi K3 and GLM-5.3 on DeepSWE, while Nous says its 33.89 Hermes Index score matches GPT-6 Luna — both are partner claims.
阶跃星辰 StepFun · 🔥 0
Experience and discussion from the community
Share my Step 5 Preview experienceAsk about Step 5 Preview
Nobody has shared their experience with Step 5 Preview yet.