Qwen3-8B places in the top three on AutoLogi.
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 | |
|---|---|---|---|---|
| Vectara Hallucination Leaderboard (2025 dataset) | 4.80% | #9 of 104 | Official leaderboard | |
| AlGhafa | 74.8% | #7 of 39 | Official leaderboard | |
| ArbMMLU-HT | 61.5% | #7 of 39 | Official leaderboard | |
| ArabicMMLU | 65% | #8 of 39 | Official leaderboard | |
| Open Arabic LLM Leaderboard v2 | 66.2% | #8 of 39 | Official leaderboard | |
| ALRAGE | 78.2% | #12 of 39 | Official leaderboard | |
| Vectara Hallucination Leaderboard (HHEM) | 3.00% | #47 of 148 | Official leaderboard | |
| MadinahQA | 52.2% | #14 of 39 | Official leaderboard | |
| Global PIQA | 74% | #17 of 44 | Official leaderboard | |
| AlignBench | 8.46 | #6 of 15 | Lab model card | |
| IFStruct | 79.8% | #4 of 10 | Official leaderboard | |
| Arena-Hard-Auto | 85.8% | #16 of 39 | Lab model card | |
| Stick to Your Role! | 0.61 | #7 of 15 | Official leaderboard | |
| RULER | 89.1% | #22 of 42 | Lab model card | |
| Fiction.LiveBench | 62.1% | #31 of 57 | Fiction.live leaderboard, via Epoch AI | |
| Multi-IF | 71.2% | #16 of 28 | Lab model card | |
| IFEval | 85% | #81 of 135 | Lab model card | |
| Phare | 63.9% | #38 of 62 | Official leaderboard | |
| MLogiQA | 69% | #5 of 8 | Lab model card | |
| Encyclo-K | 26.8% | #15 of 23 | Official leaderboard | |
| MathIF | 37.9% | #8 of 12 | Official leaderboard | |
| MMLU-Redux | 87.5% | #31 of 45 | Lab model card | |
| INCLUDE | 67.8% | #22 of 31 | Lab model card | |
| MT-AIME2024 | 65.4% | #5 of 7 | Lab model card | |
| SEA-HELM | 60.3 | #52 of 71 | Official leaderboard | |
| C-Eval | 83.4% | #35 of 46 | Lab model card | |
| MMMLU | 74.4% | #55 of 71 | Lab model card | |
| BOUQuET | 22.3 | #14 of 18 | Official leaderboard | |
| AraGen | 26.6% | #33 of 42 | Official leaderboard | |
| WritingBench | 75.9 | #11 of 14 | Lab model card | |
| Creative Writing v3 | 75.0 | #10 of 12 | Lab model card | |
| HABIT-Bench | 30.3% | #5 of 6 | Official leaderboard | |
| EQ-Bench 3 | 909 | #67 of 78 | Official leaderboard | |
| KINA | 22.3% | #32 of 37 | Official leaderboard | |
| QIMMA | 42.2% | #22 of 25 | Official leaderboard | |
| Alyah | 25.7% | #10 of 10 | Official leaderboard | |
| DiscoX | 28.4 | #7 of 7 | Official leaderboard | |
| SwitzerlandQA | 60.7% | #7 of 7 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH-500 | 97.4% | #13 of 70 | Lab model card | |
| AutoLogi | 89.1% | #3 of 8 | Lab model card | |
| ZebraLogic | 84.8% | #13 of 23 | Lab model card | |
| AIME 2024 | 76% | #51 of 88 | Lab model card | |
| OTIS Mock AIME 2024-2025 | 56.1% | #129 of 212 | Epoch AI run | |
| GPQA Diamond | 56.8% | #141 of 229 | Epoch AI run | |
| HiPhO | 31.4% | #28 of 36 | Official leaderboard | |
| PolyMath | 42.7% | #19 of 24 | Lab model card | |
| DTBench | 59.7% | #146 of 180 | Official leaderboard | |
| AIME 2025 | 67.3% | #116 of 143 | Lab model card | |
| SGI-Bench | 19.1% | #15 of 15 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AetherCode | 10% | #6 of 9 | Official leaderboard | |
| Codeforces rating | 1,785 | #32 of 47 | Lab model card | |
| LiveCodeBench v5 | 57.5% | #33 of 47 | Lab model card | |
| ArtifactsBench (Gemini 2.5 Pro 0506 judge) | 36.5% | #13 of 18 | Official leaderboard | |
| Multi-LCB | 33.9% | #8 of 11 | Official leaderboard | |
| AutoCodeBench | 28.5% | #29 of 34 | Official leaderboard | |
| GitTaskBench | 1.85% | #10 of 10 | Official leaderboard | |
| ContextBench | 4.80% | #9 of 9 | Official leaderboard | |
| ABC-Bench | 8.30% | #7 of 7 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| C3-Bench | 8.98% | #11 of 30 | Official leaderboard | |
| BFCL V4 | 42.6% | #25 of 65 | Official leaderboard | |
| SnakeBench | 25 | #79 of 188 | Official leaderboard | |
| Berkeley Function Calling Leaderboard (BFCL) | 68.1% | #23 of 41 | Lab model card | |
| PlanBench-XL | 0% | #8 of 9 | Official leaderboard | |
| LisanBench | 51.3 | #123 of 133 | Official leaderboard | |
| AdaPlanBench | 14.4% | #10 of 10 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BRIDGE | 45.5 | #22 of 39 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AraTrust | 83.4% | #10 of 39 | Official leaderboard | |
| MolecularIQ | 11.2% | #7 of 17 | Official leaderboard | |
| AthenaBench | 40.1% | #4 of 6 | Official leaderboard | |
| nocot-bench (NCRI) | 51.4 | #161 of 185 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 67.1% | #14 of 27 | Lab model card | |
| ACCoRD | 41.8 | #130 of 152 | Official leaderboard | |
| Conceptual Reasoning Index | 20.2 | #135 of 151 | Official leaderboard | |
| LMCA | 8.82% | #136 of 147 | 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 130.4 to 133.0. 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.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| GPQA Diamond | 62% GPQA-Diamond, thinking mode (Table 17). Non-thinking 39.3 (Table 18) | 56.8% Epoch AI run | +5.2 pts |
| Berkeley Function Calling Leaderboard (BFCL) | 68.1% BFCL v3, thinking mode (Table 17). Non-thinking 60.2 (Table 18) | 66.3% Leaderboard, native function calling | +1.8 pts |
| PolyMath | 42.7% PolyMath, thinking mode (Table 17). Non-thinking 18.8 (Table 18) | 42.7% Leaderboard, thinking | Same |
| RULER | 89.1% RULER Avg. (4K to 128K, YaRN x4), non-thinking mode, 128K 77.4. Thinking mode Avg. 84.4 with 8192-token thinking budget (Table 23) | 89.1% 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.