qwen3-235b-a22b: benchmark scores, price and ranking
qwen3-235b-a22b from Alibaba ranks #397 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.9), based on 8 benchmark results captured up to 2026-10-13.
- Organisation
- Alibaba
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2025-04-29 (source: curated)
- Cheapest API price
- $0.180 in / $0.540 out per million tokens (via deepinfra, observed 2026-10-13T19:47Z)
- Aggregate rank
- #397 of 743 (score 47.9)
Relative to other models, qwen3-235b-a22b is strongest on Epoch ECI (#55 of 204) and weakest on SimpleBench (#41 of 61).
Across the 8 benchmarks where at least ten models are scored, qwen3-235b-a22b is in the top 10% on 0 and in the top half on 4.
Its cheapest listed input price, $0.180 per million tokens, is lower than that of 70% of the 427 models with a listed price.
qwen3-235b-a22b benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| Epoch ECI | 139 | #55 of 204 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Fiction.liveBench | 67.7% | #13 of 47 | Epoch AI Benchmarking Hub | 2026-10-13 |
| MMLU-Pro | 68.2% | #45 of 136 | HuggingFace MMLU-Pro leaderboard | 2026-10-13 |
| DTBench | 75.7% | #65 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 70.7% | #144 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 25.0% | #64 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 37.3% | #72 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SimpleBench | 31.0% | #41 of 61 | Epoch AI Benchmarking Hub | 2026-10-13 |
About qwen3-235b-a22b
qwen3-235b-a22b is an open-weight model that can run on your own hardware from Alibaba. It was released on 2025-04-29, according to the developer's announcement. The cheapest API listing we track charges $0.180 per million input tokens and $0.540 per million output tokens, through deepinfra. We hold 8 benchmark results for it across 3 areas, and none is in the top tenth of its board.
Where qwen3-235b-a22b stands
On our aggregate score, qwen3-235b-a22b is #397 of 743 ranked models, around the middle of the table. Its score is 47.9; the scale is centred on 50, so it sits 2.1 points below the typical model. That beats 47% of the ranked field. Among the 74 ranked Alibaba models it is #41. The next one up is qwen3-4b at 47.9. The next one down is Qwen2-VL-72B-Instruct at 47.6.
qwen3-235b-a22b, benchmark by benchmark
Epoch ECI puts qwen3-235b-a22b at 139, #55 of 204. That beats 73% of the models tested, but 17 points short of Qwen 3.8 Max (156). The results either side are Qwen3.5 9B (139) and DeepSeek-R1 (139). It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so qwen3-235b-a22b is 12 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On Fiction.liveBench, qwen3-235b-a22b reaches 67.7%. It ranks #13 of 47, ahead of 72% of the field, and 26.7 points behind grok-4-0709 at 94.4%. The results either side are MiniMax-M1-80k (69.4%) and chatgpt-4o-01-29-2025 (66.7%). It covers long-context story comprehension (16k). The board median is 61.1%, so qwen3-235b-a22b is 6.6 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen3-235b-a22b manages 68.2% on MMLU-Pro, good for #45 of 136. Darwin-180B-RSI leads at 88.1%, 19.9 points further up, and 67% of models sit below. The results either side are mimo-v2.5-pro (68.5%) and Mistral-Large-Instruct-2411 (67.9%). It covers 57-subject knowledge + reasoning. The board median is 55.3%, so qwen3-235b-a22b is 12.9 points above it. Source: HuggingFace MMLU-Pro leaderboard, captured 2026-10-13.
DTBench puts qwen3-235b-a22b at 75.7%, #65 of 146. That beats 55% of the models tested, but 23.2 points short of Claude Opus 5.5 (max) (98.9%). The results either side are gpt-oss-120b (high) (76.3%) and Mistral Medium 3.5 (75.5%). It covers dynamic tool-use benchmark. The board median is 68.8%, so qwen3-235b-a22b is 6.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen3-235b-a22b lands below the middle on GPQA Diamond: 70.7%, #144 of 275, with 52% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is Mistral-Small-4-119B-2603 at 71.2%. It covers expert-level science questions. The board median is 71.7%, so qwen3-235b-a22b is 1.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen3-235b-a22b lands below the middle on LMCA: 25.0%, #64 of 108, with 58% of the field scoring higher. Claude Opus 5.5 (max) leads at 68.2%; the closest result above is GPT-4.1 at 25.6%. It covers language-model capability assessment. The board median is 29.2%, so qwen3-235b-a22b is 4.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen3-235b-a22b lands below the middle on WeirdML: 37.3%, #72 of 109, with 65% of the field scoring higher. GPT-6 Astra (max) leads at 93.3%; the closest result above is DeepSeek-V3.1 at 37.5%. It covers unusual machine-learning tasks. The board median is 41.2%, so qwen3-235b-a22b is 3.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen3-235b-a22b lands below the middle on SimpleBench: 31.0%, #41 of 61, with 66% of the field scoring higher. Claude Fable 5 (max) leads at 81.9%; the closest result above is gemini-exp-1206 at 31.1%. It covers trick questions that test common sense. The board median is 41.4%, so qwen3-235b-a22b is 10.4 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Reasoning, maths and science: uneven, ahead of 49% of the field on average across 3 benchmarks. Its best board here is MMLU-Pro (#45 of 136) and its weakest is SimpleBench (#41 of 61). General knowledge and chat: solid, ahead of 55% of the field on average across 4 benchmarks. Its best board here is Epoch ECI (#55 of 204) and its weakest is WeirdML (#72 of 109).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so qwen3-235b-a22b costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on codestral-2405 ($0.200) and DeepSeek-R1-Distill-Llama-70B ($0.200).
Release history
The Alibaba model released before it was qwen3-1.7b on 2025-04-29, which scores 46.5. The next one was Qwen3-32B on 2025-04-29, at 46.9.
Reading these numbers
The 8 scores on this page come from 2 sources: Epoch AI Benchmarking Hub (7), HuggingFace MMLU-Pro leaderboard (1). No official leaderboard supplied any of them, so treat them as third-party measurements.
Models ranked nearby
- #394 qwen3-4b (47.9)
- #395 DeepSeek-R1-Distill-Qwen-32B (47.9)
- #396 DeepSeek-V3.2 (Thinking; Fireworks) (47.9)
- #398 Opus 4.5 (Thinking, None) (47.9)
- #399 glm-4p7 (47.9)
- #400 gemini-2.0-flash-thinking-exp-01-21 (47.9)
More from Alibaba
- Qwen 3.6 Max (Preview) (#25, 54.1)
- Qwen3.8 Max (xhigh) (#37, 53.2)
- Qwen3.7 Max (#71, 52.3)
- ollama/qwen2.5-coder:32b (#93, 51.9)
- Qwen3 235B A22B diff, no think, Alibaba API (#122, 51.5)
- Qwen2.5-Coder-14B-Instruct (#135, 51.3)
Compare it on the live leaderboard → · How the aggregate score works