Model Details
Step3-VL-10B
Quick answer: Step3-VL-10B is StepFun's 10B-parameter vision-language model, scoring 84.0% on MathVista's visual math reasoning benchmark.
At a Glance
Where Step3-VL-10B leads
- 84.0% MathVista — strong visual math reasoning for its parameter size
Where it lags
- Proprietary: no open weights
- Fewer independently-verified benchmark results than larger frontier models
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.