Step 3.5 Flash places in the top three on 5 benchmarks.
StepFun Feb 2026 Open weights Mid-size
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.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMLU-Pro | 84.4% | #37 of 151 | Model card | |
| Arena-Hard v2.0 | 74% | #9 of 24 | Model card | |
| LMArena (Elo) | 1,393 | #137 of 333 | Official leaderboard | |
| RepoQA | 88.5% | #2 of 4 | Model card | |
| BullshitBench | 25.4% | #58 of 107 | Official leaderboard | |
| LongBench v2 | 57.5% | #15 of 27 | Model card | |
| MultiChallenge | 55.7% | #21 of 36 | Model card | |
| Humanity's Last Exam | 23.1% | #90 of 147 | Model card | |
| BullshitBench V2 | 22% | #80 of 122 | Official leaderboard | |
| IFBench | 67.4% | #48 of 70 | Model card | |
| Extended NYT Connections | 28.4% | #75 of 97 | Official leaderboard | |
| FRAMES | 76.5% | #7 of 8 | Model card | |
| OpenAI MRCR | 28.8% | #10 of 10 | Model card | |
| Consulting and Recommendations Benchmark (StepFun) | 70.5% | #1 of 1 | Model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BRUMO 2025 | 100% | #2 of 42 | Official leaderboard | |
| CMIMC 2025 | 95.6% | #3 of 33 | Official leaderboard | |
| HMMT Nov 2025 | 94.2% | #2 of 22 | Official leaderboard | |
| HMMT 2025 | 98.4% | #5 of 49 | Model card | |
| AIME 2025 | 97.3% | #16 of 143 | Model card | |
| SMT 2025 | 91.5% | #5 of 40 | Official leaderboard | |
| AIME 2026 | 96.7% | #7 of 31 | Official leaderboard | |
| GPQA Diamond | 83.5% | #86 of 284 | Model card | |
| IMO-AnswerBench | 85.4% | #13 of 37 | Model card | |
| MathArena (Apex, IMO, USAMO) | 13.5% | #19 of 47 | Official leaderboard | |
| HMMT 2026 | 86.4% | #22 of 31 | Official leaderboard | |
| USAMO 2026 | 44.6% | #7 of 9 | Official leaderboard | |
| ARC-AGI | 54.8% | #15 of 19 | Model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 86.4% | #17 of 76 | Model card | |
| SWE-bench Verified | 74.4% | #49 of 150 | Model card | |
| SWE-bench Multilingual | 67.4% | #46 of 70 | Model card | |
| CF-Div2-Stepfun | 86.1% | #1 of 1 | Model card | |
| Data Analysis Benchmark (StepFun) | 39.6% | #1 of 1 | Model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BrowseComp-ZH | 73.7% | #4 of 17 | Model card | |
| tau2-bench | 88.2% | #18 of 69 | Model card | |
| LisanBench | 2409.7 | #39 of 133 | Official leaderboard | |
| PinchBench 2.0 | 79.3% | #18 of 57 | Official leaderboard | |
| xbench-DeepSearch | 83.7% | #1 of 3 | Model card | |
| GAIA | 84.5% | #2 of 5 | Model card | |
| PinchBench | 70.5% | #22 of 48 | Official leaderboard | |
| SnakeBench | 23 | #97 of 183 | Official leaderboard | |
| BrowseComp | 69% | #44 of 77 | Model card | |
| ResearchRubrics | 65.3% | #3 of 5 | Model card | |
| Terminal-Bench 2.0 | 51% | #37 of 61 | Model card | |
| AndroidDaily | 57% | #2 of 2 | Model card | |
| xbench-DeepSearch-2510 | 56.3% | #1 of 1 | Model card |
The benchmarks the lab chose to headline, with the scores it gave.
Every result in the longer documents the lab published with it. 16 of them aren't in the launch post.