Step releases weights for Step 5 Preview model
10/15/2026 — 10/15, 12:08·1 sources (all)·1 reports (all)·In the 2026-10-15 briefing
Key takeaways
- Step officially released weights for its Step 5 Preview model, including BF16 and FP8 versions, technical blog and open-source tools on Hugg…
Background & analysis
On October 15, 2026, Step officially released the weights for Step 5 Preview. The release includes BF16 and FP8 model weights, a technical blog, and accompanying open-source tools. Step 5 Preview features a sparse MoE architecture with 600B total parameters and 27B activated parameters per token. It supports a 1M token context window and natively handles both text and visual inputs. Designed for long-context Agent tasks, the model excels in scenarios like AI programming, software engineering, professional knowledge work, and finance. According to the company, Step 5 Preview recently ranked second globally in both weekly and daily call volume on OpenRouter. The model weights are now available on Hugging Face and ModelScope, with support for various hardware platforms and inference frameworks. Developers can deploy API services compatible with OpenAI's interface using vLLM or SGLang, or perform offline inference with Transformers.
AI-generated from 1 reports · updated 55 minutes ago
Latest turnStep has officially released the Step 5 Preview model weights in BF16 and FP8, along with a technical blog and open-source tooling, now available on Hugging Face and ModelScope. The sparse MoE model has 600B total parameters with 27B active per token, a 1M token context window, native text and vision input, and is designed for long-horizon agent tasks. Developers can deploy it with vLLM or SGLang, and Step says it ranks second globally in OpenRouter weekly usage.
- Heat index
- 95
- All sources
- 1
- All reports
- 1
- First seen
- 2 hours ago
Reports on this story headlines open the original
Step has officially released the Step 5 Preview model weights in BF16 and FP8, along with a technical blog and open-source tooling, now available on Hugging Face and ModelScope. The sparse MoE model has 600B total parameters with 27B active per token, a 1M token context window, native text and vision input, and is designed for long-horizon agent tasks. Developers can deploy it with vLLM or SGLang, and Step says it ranks second globally in OpenRouter weekly usage.
IT之家AI score 68
Other stories people are talking about
- Heat index 51WeChat Work and WorkBuddy Deeply Integrate with Step 5 Preview1 sources
- Heat index 735Google rolls out Gemini 4 Argon globally18 sources
- Heat index 360Meta launches Muse Wonder and Muse Wonder Full models11 sources
- Heat index 346Anthropic Study Warns Small AI Models Could Spread Across Devices10 sources
- Heat index 346Google Play Launches Personal AI App Execution Feature9 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion