Step Edge

StepFun
Model overview

Step Edge

Step Edge is a text-and-vision base model for phones, vehicles, and other edge scenarios, covering image understanding, GUI grounding, tool use, and spatial and video understanding.

multimodalvision-languageagentcomputer useedge / on-devicelocal inference
Representative version

Step Edge

CompanyStepFun
Release2026-07-12
Release typeModel release
ParametersNot disclosed
ArchitectureNot disclosed
ContextNot disclosed
Input modalitiestext / image / video
Output modalitiestext / tool_call
Tool use / function callingConfirmed: Tool use
Structured output / reasoningStructured output: Not disclosed; Reasoning: Not disclosed
Vision / audio / videoConfirmed: Vision, Video
Open weights / LicenseWeight status not disclosed / license not disclosed
stepfun-step-edge-current · License source pending verification
Vendor APINot disclosed
API source pending verification
OpenRouterNot separately listed on OpenRouter
PricingNo reliable public source
BenchmarkNo reliable public source
Public signalsNo reliable public source
Use casesEdge assistants / In-vehicle interaction / Local automation / local inference
Claim evidenceRelease event / Model specificationsStep Edge | Efficient Multimodal Models for Real-World Edge ScenariosStep Edge · 2026-07-12 · Step Edge · Step Edge
OpenRouter

Pricing / context / availability

OpenRouterNot separately listed on OpenRouter
OpenRouter listingNot separately listed in the current OpenRouter public list
Mapped versionStep Edge
stepfun-step-edge-current
PricingNo reliable public source
ContextNot disclosed
Speed/LatencyNo reliable public source
Snapshot time2026-08-11T06:28:09.895Z
SourceView source
Verified limitations

Usage boundaries

  • The reviewed release page does not disclose a parameter count, context length, public API, open weights, or license for Step Edge.
Missing public information

Public information not yet available

Parameter scale not disclosedNot listed on OpenRouterNo reliable benchmark sourceNo public model-card activity countersNo linkable model-specific reviewSpeed/latency not disclosed