编程能力榜
按 Artificial Analysis · 编程专项 口径排列的 25 款大模型,Claude Fable 5.2、Claude Opus 5.5、Gemini 4 Carbon 领跑榜单。数据聚合自公开评测源,更新于 2026-11-21。
口径:比较代码生成、调试与工程任务表现,面向开发者的选型口径。,分数越高越好。评测任务成本不是 API Token 单价;不同来源的分数不能直接相加或横向比较。
编程能力榜
| 排名 | 模型 | 编程分 | 上下文 | 输入 /M | 输出 /M | 详情 |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2Anthropic | 64 | 1M | $5 | $25 | |
| 2 | Claude Opus 5.5Anthropic | 62 | 1M | $4 | $20 | |
| 3 | Gemini 4 CarbonGoogle | 58 | 1M | $2.50 | $12 | |
| 4 | GPT-6 AstraOpenAI | 56 | 1M | $2.50 | $10 | |
| 5 | Claude Sonnet 5.5Anthropic | 54 | 1M | $3 | $15 | |
| 6 | Claude Fable 5.1Anthropic | 52 | 1M | $4 | $16 | |
| 7 | GPT-6.1 SolOpenAI | 51 | 1M | $0.60 | $2.40 | |
| 8 | Gemini 4 ArgonGoogle | 50 | 1M | $1.25 | $5 | |
| 9 | MiMo-V2.6-Pro小米 | 48 | 1M | $0.30 | $1.20 | |
| 10 | GLM-5.3智谱 | 48 | 1M | $0.80 | $3.20 | |
| 11 | Kimi K3月之暗面 | 46 | 1.05M | $0.70 | $2.80 | |
| 12 | Muse Spark 1.3Meta | 45 | 1M | $1 | $4 | |
| 13 | Grok 4.7xAI | 45 | 500K | $3 | $12 | |
| 14 | Qwen3.8 Max阿里巴巴 | 45 | 984K | $1.60 | $6.40 | |
| 15 | DeepSeek V4深度求索 | 45 | 128K | $0.28 | $1.10 | |
| 16 | Step 5 Preview阶跃星辰 | 43 | 1M | $0.60 | $2.40 | |
| 17 | GLM-5.3-Flash智谱 | 43 | 1M | $0.25 | $1 | |
| 18 | Claude Haiku 5.5Anthropic | 41 | 1M | $1 | $5 | |
| 19 | GPT-5.6 TerraOpenAI | 41 | 1M | $0.90 | $3.60 | |
| 20 | Qwen3.8 2.4T A95B阿里巴巴 | 41 | 262K | $0.90 | $3.60 | |
| 21 | Ling 3.1 Flash蚂蚁集团 | 40 | 1M | $0.40 | $1.60 | |
| 22 | Qwen3.8-Flash-Next阿里巴巴 | 39 | 256K | $0.15 | $0.60 | |
| 23 | Gemini 3.8 FlashGoogle | 38 | 1M | $0.50 | $2 | |
| 24 | Mercury 2.5Inception Labs | 38 | 128K | $0.50 | $1.50 | |
| 25 | Celeris-1Inception Labs | 36 | 128K | $0.70 | $2.10 |
切换榜单视角
返回模型排行榜,查看全部榜单视角与收录说明。