StepFun model product
Step3 VL 10B is a lightweight open-source multimodal foundation model with a language-aligned Perception Encoder and Qwen3-8B decoder.
Updated Sep 8, 2026. Default version: Step3-VL-10B
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for Step3-VL-10B.
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkMMBench | Score91.80% | Rank01 | Participants9 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkMathVista | Score84.00% | Rank05 | Participants39 | Percentile89.47% | EvidenceC | Evaluated |
| BenchmarkMulti-Challenge | Score62.60% | Rank07 | Participants29 | Percentile78.57% | EvidenceC | Evaluated |
| BenchmarkMMMU | Score78.10% | Rank17 | Participants64 | Percentile74.60% | EvidenceC | Evaluated |
| BenchmarkMathVision | Score70.80% | Rank18 | Participants35 | Percentile50.00% | EvidenceC | Evaluated |
| BenchmarkAIME 2025 | Score87.70% | Rank56 | Participants119 | Percentile53.39% | EvidenceC | Evaluated |
Preference and agent-evaluation results for the default version.
The default version does not have a matching Arena result yet.
Official vendor API pricing appears first, followed by individual provider offers.
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Provider-specific output speed and catalog latency for Step3-VL-10B. Runtime does not affect the capability score.
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Technical details for the model's default version.
Available versions of this model. The score column identifies the version used in the overall ranking.
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionStep3-VL-10B | Released | LLMBoard42.55 | Parameters10B | ContextN/A | Max outputN/A | Open weightsYes | LicenseApache 2.0 |
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Key information about Step3 VL 10B and its available data.
2T multimodal tokens using a language-aligned Perception Encoder and Qwen3-8B decoder. It includes Parallel Coordinated Reasoning (PaCoRe) for scaling test-time compute for complex perceptual reasoning.
Data as of 2026-09-08.
Common questions about Step3 VL 10B.
Step3 VL 10B's default version was released on Jan 15, 2026.
No official standard PAYG price is currently available for Step3 VL 10B.
Step3 VL 10B was created by StepFun.
A context window is not available for the default version.
Yes. The default version is marked as open weight under Apache 2.0.
No provider offering is currently linked to the default version.