Model

Step-1V

CLOSED

StepFun · Step-1 family · released Sep 1, 2024

StepFun's first publicly released multimodal (vision) model, adding image understanding atop the Step-1 text base in September 2024; since deprecated in favor of newer Step vision models.

ReasoningCodingVisionFunction callingTool useAgentic

1978.9

Elo · unrated

Parameters

Undisclosed

Active params

—

Context

32K tokens

Architecture

Dense multimodal transformer (LLM backbone + vision encoder), exact scale undisclosed

License

Proprietary

Languages

—

API price (in/out)

No hosted API

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

OCRBenchVision886#3

best: InternVL3-78B · 906

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

—

Output / M tok

—

Step-1 family

Elo progression across releases

Step-1Mar 20241000.0 Step-1VSep 20241978.9

Resources

DOCSplatform.stepfun.com/docs/api-reference/models/list

API price — · each benchmark row carries its own source badge (see methodology)