Qwen3 235B A22B
Qwen3 235B A22B is a large language model developed by Alibaba, featuring a Mixture-of-Experts (MoE) architecture with 235 billion total parameters and 22 billion activated parameters. It achieves competitive results in benchmark evaluations of coding, math, general capabilities, and more, compared to other top-tier models.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| Aider | 61.8% | self-reported llm-stats | link → |
| AIME 2024 | 85.7% | self-reported llm-stats | link → |
| AIME 2025 | 81.5% | self-reported llm-stats | link → |
| Arena Hard | 95.6% | self-reported llm-stats | link → |
| BBH | 88.9% | self-reported llm-stats | link → |
| BFCL | 70.8% | self-reported llm-stats | link → |
| CRUX-O | 79.0% | self-reported llm-stats | link → |
| EvalPlus | 77.6% | self-reported llm-stats | link → |
| GPQA | 47.5% | self-reported llm-stats | link → |
| GSM8k | 94.4% | self-reported llm-stats | link → |
| Include | 73.5% | self-reported llm-stats | link → |
| LiveBench | 77.1% | self-reported llm-stats | link → |
| LiveCodeBench | 70.7% | self-reported llm-stats | link → |
| MATH | 71.8% | self-reported llm-stats | link → |
| MBPP | 81.4% | self-reported llm-stats | link → |
| MGSM | 83.5% | self-reported llm-stats | link → |
| MMLU | 87.8% | self-reported llm-stats | link → |
| MMLU-Pro | 68.2% | self-reported llm-stats | link → |
| MMLU-Redux | 87.4% | self-reported llm-stats | link → |
| MMMLU | 86.7% | self-reported llm-stats | link → |
| MultiLF | 71.9% | self-reported llm-stats | link → |
| MultiPL-E | 65.9% | self-reported llm-stats | link → |
| SuperGPQA | 44.1% | self-reported llm-stats | link → |