Qwen3-30B-A3B places in the top three on MT-AIME2024.
Qwen Apr 2025 Open weights
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 | |
|---|---|---|---|---|
| LLM Confabulation Leaderboard | 12.3% | #7 of 58 | Official leaderboard | |
| Arena-Hard-Auto | 91% | #9 of 39 | Lab model card | |
| AlignBench | 8.70 | #4 of 15 | Lab model card | |
| RULER | 91.6% | #16 of 42 | Lab model card | |
| MT-AIME2024 | 73.9% | #3 of 7 | Lab model card | |
| Encyclo-K | 31.6% | #11 of 23 | Official leaderboard | |
| IFEval | 86.5% | #67 of 135 | Lab model card | |
| Multi-IF | 72.2% | #14 of 28 | Lab model card | |
| MLogiQA | 70.1% | #4 of 8 | Lab model card | |
| MMLU-Redux | 89.5% | #26 of 45 | Lab model card | |
| LMArena (Elo) | 1,326 | #201 of 335 | Official leaderboard | |
| INCLUDE | 71.9% | #19 of 31 | Lab model card | |
| C-Eval | 86.6% | #29 of 46 | Lab model card | |
| Arena-Hard v2.0 | 33.9% | #15 of 23 | Official leaderboard | |
| MMMLU | 78.4% | #47 of 71 | Lab model card | |
| LongBench v2 | 42.5% | #22 of 33 | Official leaderboard | |
| WritingBench | 77.0 | #10 of 14 | Lab model card | |
| Creative Writing v3 | 79.1 | #9 of 12 | Lab model card | |
| KINA | 24.3% | #30 of 37 | Official leaderboard | |
| EQ-Bench 3 | 917 | #66 of 78 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 35.6 | #105 of 119 | Official leaderboard | |
| Fiction.LiveBench | 40.6% | #52 of 57 | Fiction.live leaderboard, via Epoch AI |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH-500 | 98% | #8 of 70 | Lab model card | |
| ZebraLogic | 89.5% | #6 of 23 | Lab model card | |
| AIME 2024 | 80.4% | #41 of 88 | Lab model card | |
| Misguided Attention | 43.6% | #24 of 45 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 62.8% | #122 of 212 | Epoch AI run | |
| GPQA Diamond | 61.7% | #135 of 229 | Epoch AI run | |
| HiPhO | 47.1% | #22 of 36 | Official leaderboard | |
| AutoLogi | 88.7% | #5 of 8 | Lab model card | |
| PolyMath | 46.1% | #17 of 24 | Lab model card | |
| HMMT 2025 | 50.8% | #38 of 53 | Official leaderboard | |
| AIME 2025 | 70.9% | #111 of 143 | Lab model card | |
| DTBench | 60.3% | #143 of 180 | Official leaderboard | |
| SMT 2025 | 67.9% | #35 of 40 | Official leaderboard | |
| BRUMO 2025 | 77.5% | #38 of 42 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v5 | 62.6% | #23 of 47 | Lab model card | |
| Multi-LCB | 45.5% | #6 of 11 | Official leaderboard | |
| Codeforces rating | 1,974 | #28 of 47 | Lab model card | |
| WeirdML | 29.8% | #99 of 112 | Official leaderboard | |
| DesignArena Website | 948 | #139 of 151 | Official leaderboard | |
| DesignArena UI Component | 937 | #129 of 139 | Official leaderboard | |
| DesignArena Code | 932 | #137 of 143 | Official leaderboard | |
| DesignArena Game Dev | 889 | #139 of 143 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| C3-Bench | 9.77% | #7 of 30 | Official leaderboard | |
| Berkeley Function Calling Leaderboard (BFCL) | 69.1% | #20 of 41 | Lab model card | |
| Step Race | 2 | #32 of 59 | Official leaderboard | |
| SnakeBench | 16 | #167 of 188 | Official leaderboard | |
| Elimination Game | 2 | #55 of 56 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BRIDGE | 47.4 | #16 of 39 | Official leaderboard | |
| FinEval 6.0 | 79.9% | #2 of 4 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 50.4 | #164 of 185 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 74.3% | #9 of 27 | Lab model card | |
| ACCoRD | 45.0 | #123 of 152 | Official leaderboard | |
| Conceptual Reasoning Index | 25.2 | #126 of 151 | Official leaderboard | |
| LMCA | 15.8% | #123 of 147 | Official leaderboard | |
| Political Science LLM Benchmark | 59.3% | #42 of 42 | Official leaderboard |
The Self-Reported Index fits one score to the 18 capability results Qwen 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 132.4 to 135.4. 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.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| GPQA Diamond | 65.8% GPQA-Diamond, thinking mode (Table 15). Non-thinking 54.8 (Table 16) | 61.7% Epoch AI run | +4.1 pts |
| Berkeley Function Calling Leaderboard (BFCL) | 69.1% BFCL v3, thinking mode (Table 15). Non-thinking 58.6 (Table 16) | 65.7% Leaderboard, native function calling | +3.4 pts |
| AIME 2025 | 70.9% AIME’25, thinking mode (Table 15). Non-thinking 21.6 (Table 16) | 70% Leaderboard | +0.9 pts |
| PolyMath | 46.1% PolyMath, thinking mode (Table 15). Non-thinking 23.3 (Table 16) | 46.1% Leaderboard, thinking | Same |
| RULER | 91.6% RULER Avg. (4K to 128K, YaRN x4), non-thinking mode, 128K 79.2. Thinking mode Avg. 86.6 with 8192-token thinking budget (Table 23) | 91.6% Leaderboard | Same |
Qwen's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
Every result in the longer documents the lab published with it. 24 of them aren't in the launch post.