Model card
Qwen3-235B-A22B.
Alibabaopen-weights235B (22B active) paramsmoe1 current SOTA
Qwen3 flagship MoE model, May 2025
§ 02 · Benchmarks
Every benchmark Qwen3-235B-A22B has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | WebArena | Agentic AI · Web & Desktop Agents | accuracy | 95.6% | #1 | — | source ↗ |
| 02 | MBPP | Computer Code · Code Generation | pass-1 | 81.4% | #2 | — | source ↗ |
| 03 | MBPP+ | Computer Code · Code Generation | pass-1 | 81.4% | #2 | — | source ↗ |
| 04 | BIG-Bench Hard | Reasoning · Multi-step Reasoning | accuracy | 88.9% | #3 | — | source ↗ |
| 05 | AIME 2024 | Reasoning · Mathematical Reasoning | accuracy | 85.7% | #6 | — | source ↗ |
| 06 | LiveCodeBench Pro | Computer Code · Code Generation | elo | 1673.00 | #6 | — | source ↗ |
| 07 | LiveCodeBench | Computer Code · Code Generation | pass@1 | 70.7% | #8 | — | source ↗ |
| 08 | AIME 2025 | Reasoning · Mathematical Reasoning | accuracy | 81.5% | #15 | — | source ↗ |
| 09 | LiveCodeBench | Computer Code · Code Generation | pass-1 | 70.7% | #15 | — | source ↗ |
| 10 | MMLU | Reasoning · Commonsense Reasoning | accuracy | 87.8% | #27 | — | source ↗ |
| 11 | GSM8K | Reasoning · Mathematical Reasoning | accuracy | 94.4% | #29 | — | source ↗ |
| 12 | MATH | Reasoning · Mathematical Reasoning | accuracy | 71.8% | #34 | — | source ↗ |
| 13 | GPQA Diamond | Reasoning · Multi-step Reasoning | accuracy | 71.1% | #47 | — | source ↗ |
| 14 | GPQA Diamond | Reasoning · Multi-step Reasoning | accuracy | 71.1% | #47 | — | source ↗ |
| 15 | PLCC | Natural Language Processing · Polish Cultural Competency | grammar | 66.0% | #61 | — | source ↗ |
| 16 | PLCC | Natural Language Processing · Polish Cultural Competency | geography | 69.0% | #93 | — | source ↗ |
| 17 | PLCC | Natural Language Processing · Polish Cultural Competency | history | 70.0% | #94 | — | source ↗ |
| 18 | PLCC | Natural Language Processing · Polish Cultural Competency | average | 55.0% | #103 | — | source ↗ |
| 19 | PLCC | Natural Language Processing · Polish Cultural Competency | vocabulary | 43.0% | #110 | — | source ↗ |
| 20 | PLCC | Natural Language Processing · Polish Cultural Competency | culture-and-tradition | 45.0% | #118 | — | source ↗ |
| 21 | PLCC | Natural Language Processing · Polish Cultural Competency | art-and-entertainment | 37.0% | #124 | — | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Strengths by area
Where Qwen3-235B-A22B actually performs.
§ 05 · Related models
Other Alibaba models scored on Codesota.
§ 06 · Sources & freshness
Where these numbers come from.
pwc-dump
11
results
sdadas/PLCC
7
results
livecodebench-pro-official
1
result
arxiv-2505.09388
1
result
qwen-model-card
1
result
9 of 21 rows marked verified.