StepFun Releases StepEdge On-Device Models
新智元 · wechat · 2026-07-12
StepFun has released the StepEdge on-device model family, targeting local execution scenarios for devices like smartphones and car infotainment systems. It covers capabilities such as text and visual understanding, voice/audio interaction, screen and GUI operations, and image generation/editing.
The article emphasizes that this is not a single model, but part of a device-cloud collaborative system:
- StepEdge: Text and visual understanding
- StepEdgeAudio: Voice and audio interaction
- StepEdgeGUI: Screen understanding, widget localization, and UI operation
- StepEdgeGen: On-device image generation and editing
The text provides multiple benchmark results, claiming top rankings among its peers in comprehensive benchmarks, OSWorld, and audio understanding metrics. It is paired with the self-developed StepInferenceNPU inference engine, highlighting local low latency, offline availability, and privacy protection. The article positions this as the final puzzle piece in StepFun's comprehensive "Pro + Flash + Edge" model layout.
Related event: StepFun Releases Step Edge On-Device Models(3 posts)→
More from Models
- Local benchmark: Flint matches Qwen3.6 quality at 50/52 while running 3.6x faster on M4 Pro — RichardSocher · 2026-10-27
- ICML 2026 position paper: LLM 'thinking traces' are a harmful anthropomorphism — rao2z · 2026-10-27
- One stray 'u' and Claude starts spelling like a Brit — cantrell · 2026-10-27
- Cohere's Invent-A-Dataset paper: zero-seed synthetic data beats frontier APIs on quality and diversity — idanbeck · 2026-10-27
- ARC Prize redesigns leaderboard: highest reasoning effort by default, CSV export — mhmazur · 2026-10-27
- Cohere: open-weight Qwen3.8-Flash-Next tried uploading probe file to Beijing cloud bucket in sandbox — wstone_bd · 2026-10-27