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.
发布日期2025年4月29日
参数规模235B
上下文长度131K
许可证Apache 2.0
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| Arena Hard | general reasoning creativity | 95.6 | 来源 |
| GSM8k | math reasoning | 94.4 | 来源 |
| BBH | reasoning math language | 88.9 | 来源 |
| MMLU | general reasoning language math | 87.8 | 来源 |
| MMLU-Redux | language reasoning math general | 87.4 | 来源 |
| MMMLU | language reasoning math general | 86.7 | 来源 |
| AIME 2024 | math reasoning | 85.7 | 来源 |
| MGSM | math reasoning | 83.5 | 来源 |
| AIME 2025 | math reasoning | 81.5 | 来源 |
| MBPP | reasoning general | 81.4 | 来源 |
| CRUX-O | reasoning | 79.0 | 来源 |
| EvalPlus | reasoning code | 77.6 | 来源 |
| LiveBench | math reasoning general | 77.1 | 来源 |
| Include | general | 73.5 | 来源 |
| MultiLF | general | 71.9 | 来源 |
| MATH | math reasoning | 71.8 | 来源 |
| BFCL | general reasoning | 70.8 | 来源 |
| LiveCodeBench | reasoning general code | 70.7 | 来源 |
| MMLU-Pro | language reasoning math general | 68.2 | 来源 |
| MultiPL-E | general language | 65.9 | 来源 |
| Aider | reasoning code | 61.8 | 来源 |
| GPQA | reasoning general | 47.5 | 来源 |
| SuperGPQA | reasoning general math legal healthcare finance chemistry economics physics | 44.1 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| Alibaba (China) | $0.29 | $1.15 | 131K | — | — | ✓ | ✗ | ✗ |
| 302.AI | $0.29 | $2.86 | 128K | — | — | ✓ | ✗ | ✗ |
| Alibaba | $0.70 | $2.80 | 131K | — | — | ✓ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。