Step3 model card (Hugging Face) 31 Jul 2025
- MMMU 74.2
- MATH-Vision 64.8
- SimpleVQA 62.2
- Hallusion Bench 64.2
- ZeroBench (sub) 23
- DYNMATH 50.1
- AIME25 82.9
- GPQA-Diamond 73
- LiveCodeBench (202408-202505) 67.1
- HMMT25 70
- CNMO24 83.7
Step 3
Step 3 places in the top three on CNMO 2024.
15
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
StepFun Jul 2025 Step 3
Open weights Flagship
No Epoch Capabilities Index score yet
- SimpleVQA 62.2% #16 /23
- HallusionBench 64.2% #11 /23
- AIME 2025 82.9% #83 /143
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HallusionBench | 64.2% | #11 of 23 | Model card | |
| LMArena (Elo) | 1,349 | #177 of 333 | Official leaderboard | |
| SimpleVQA | 62.2% | #16 of 23 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CNMO 2024 | 83.7% | #3 of 6 | Model card | |
| GPQA Diamond | 73% | #153 of 284 | Model card | |
| AIME 2025 | 82.9% | #83 of 143 | Model card | |
| HMMT 2025 | 70% | #44 of 49 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 67.1% | #56 of 76 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 18 | #152 of 183 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMMU | 74.2% | #45 of 97 | Model card | |
| ZeroBench | 23% | #14 of 25 | Model card | |
| MATH-Vision (MathVision) | 64.8% | #28 of 42 | Model card | |
| Vision Arena | 1,145 | #103 of 136 | Official leaderboard | |
| DynaMath | 50.1% | #14 of 17 | Model card | |
| BabyVision | 14.7% | #26 of 27 | Official leaderboard |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.