Today for AI
HOT RADAR
llmHEAT 11.8°

AI CLUSTERED EVENT · 10/15/2026

StepFun Releases Step 5 Preview: 600B MoE Model with 1M Context for Long-Horizon Agents

1 reports archived1 independent sourcesupdated 10/15/2026, 11:22:04
Synthesis & Latest Updates
1 Sources Cross-Validated

StepFun has open-sourced the weights of Step 5 Preview, a 600B-parameter sparse MoE model (27B active) supporting 1M-token context and multimodal inputs. Key innovations include a 92-layer narrow-deep architecture, Sparse GQA with block-level indexing to reduce computational overhead, and RL strategies optimized for long-horizon agent tasks, positioning it as a top performer on OpenRouter.

LATEST/StepFun has open-sourced the weights of Step 5 Preview, a 600B-parameter sparse MoE model (27B active) supporting 1M-token context and multimodal inputs. Key innovations include a 92-layer narrow-deep architecture, Sparse GQA with block-level indexing to reduce computational overhead, and RL strategies optimized for long-horizon agent tasks, positioning it as a top performer on OpenRouter.

HEAT TRENDHourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. IT之家 · 智能时代T2·86 pts
    • Utilizes a sparse MoE architecture with 600B total and 27B active parameters to balance performance and cost.
    • Employs Sparse GQA with 8-token block indexing to reduce attention computation overhead by 8x for million-token contexts.
    • Integrates context compression into end-to-end RL policy training to enhance state management for long-horizon agent tasks.