StepFun
What benchmarks does StepFun mention most?
Here are the 12 benchmarks StepFun mentions most in its launch posts.
Source: Labs’ launch posts, read by Benchmark Almanac
Source: Labs’ system cards and technical reports, read by Benchmark Almanac
AIME 2025 HMMT 2025 GPQA Diamond LiveCodeBench v6 MMMU MATH-Vision (MathVision) SimpleVQA HallusionBench ZeroBench DynaMath CNMO 2024 MathVista PhyX MMBench MMStar ReMI OCRBench AI2D ScreenSpot-v2 ScreenSpot-Pro OSWorld-G BLINK All-Angles Bench Humanity's Last Exam IMO-AnswerBench BrowseComp ResearchRubrics SWE-bench Verified
Source: Labs’ launch posts, system cards and technical reports, read by Benchmark Almanac
Model cards
| Date | Model | Document | Benchmarks |
|---|---|---|---|
| Sep 20, 2026 | | Step 5 Preview: Advancing the Pareto Frontier (model page) | 33 |
| May 29, 2026 | | Step 3.7 Flash: The new frontier is agent efficiency (blog) | 20 |
| Feb 2, 2026 | | Step 3.5 Flash model card (Hugging Face) | 14 |
| Jan 13, 2026 | | STEP3-VL-10B model card (Hugging Face) | 18 |
| Jul 31, 2025 | | Step3 model card (Hugging Face) | 11 |
Every StepFun model
-
Step 5 Preview Sep 2026
-
Step 3.7 Flash May 2026
-
Step 3.5 Flash Feb 2026
-
STEP3-VL-10B Jan 2026
-
Step 3 Jul 2025
-
Step-Audio-Chat Feb 2025
-
Step1X-Edit