Hunyuan-TurboS ranks first on CMMLU.
Tencent Feb 2025 Closed weights Flagship
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 | |
|---|---|---|---|---|
| BIG-Bench Hard | 92.2% | #4 of 73 | Lab model card | |
| CMMLU | 90.8% | #1 of 17 | Lab model card | |
| DROP | 91.5% | #5 of 41 | Lab model card | |
| MMLU | 89.5% | #18 of 131 | Lab model card | |
| C-Eval | 90.9% | #14 of 46 | Lab model card | |
| Arena-Hard-Auto | 88.6% | #12 of 39 | Lab model card | |
| IFEval | 88.6% | #50 of 135 | Lab model card | |
| LMArena (Elo) | 1,383 | #151 of 335 | Official leaderboard | |
| MMLU-Pro | 79% | #87 of 167 | Lab model card | |
| Chinese SimpleQA | 70.8% | #7 of 13 | Lab model card | |
| SimpleQA | 22.8% | #36 of 60 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH | 89.7% | #11 of 99 | Lab model card | |
| GPQA Diamond | 57.5% | #226 of 302 | Lab model card | |
| AIME 2024 | 43.3% | #77 of 88 | Lab model card | |
| ZebraLogic | 46% | #23 of 23 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HumanEval | 91% | #11 of 101 | Lab model card | |
| ArtifactsBench (Gemini 2.5 Pro 0506 judge) | 51% | #3 of 18 | Official leaderboard | |
| AutoCodeBench | 43.8% | #18 of 34 | Official leaderboard | |
| LiveCodeBench | 32% | #42 of 54 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 61% | #16 of 27 | Lab model card |
The Self-Reported Index fits one score to the 16 capability results Tencent reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 124.1 to 130.0. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
The benchmarks the lab chose to headline, with the scores it gave.