编程 Agent 跑分
Codex 雷达用同一套真实编程任务给出每个模型的 IQ、费用和用时。
Codex 雷达实测 · codex
| 模型 | 推理档位 | 雷达 IQ | 平均费用(美元) | 平均用时(分钟) |
|---|---|---|---|---|
| Claude Sonnet 5 | high | 150.0 | $4.51 | 20.9 |
| Kimi K2.8 Preview | high | 150.0 | — | 14.7 |
| Kimi K2.8 Preview | max | 150.0 | — | 16.0 |
| Claude Opus 5.5 | low | 150.0 | $0.554 | 3.5 |
| Claude Opus 5.5 | medium | 150.0 | $0.665 | 3.3 |
| Claude Opus 5.5 | xhigh | 150.0 | $2.61 | 11.0 |
| Claude Opus 5.5 | max | 150.0 | $22 | 84.3 |
| Claude Opus 5 | medium | 112.5 | $5.01 | 15.1 |
| GPT-6 Astra | high | 110.4 | $2.81 | 13.8 |
| GPT-6 Astra | ultra | 110.2 | $8.75 | 15.1 |
| GPT 5.6 Sol | max | 108.5 | $6.34 | 33.9 |
| Grok 4.7 | xhigh | 108.3 | $13 | 47.0 |
| GPT-6 Astra | max | 108.0 | $4.64 | 23.3 |
| GPT-6 Astra | medium | 107.5 | $1.83 | 9.0 |
| Grok 4.6 | xhigh | 107.1 | $3.25 | 34.3 |
| GPT-6 Astra | xhigh | 105.3 | $3.60 | 17.9 |
| Grok 4.7 | medium | 105.3 | $5.92 | 35.4 |
| Grok 4.7 | high | 105.1 | $9.36 | 42.4 |
| GPT 5.6 Luna | max | 102.2 | $0.426 | 37.6 |
| Grok 4.7 | low | 101.3 | $2.76 | 19.1 |
| GPT 5.6 Terra | max | 100.9 | $2.80 | 30.9 |
| K3 | max | 100.8 | $3.33 | 59.1 |
| GPT 5.6 Sol | ultra | 100.5 | $18 | 51.9 |
| Claude Opus 5 | low | 100.0 | $1.42 | 10.7 |
| Grok 4.6 | high | 99.5 | $2.71 | 30.2 |
| Hy4 Preview | high | 99.5 | $0.656 | 38.8 |
| GPT 5.6 Terra | ultra | 99.1 | $7.65 | 41.9 |
| GPT-6 Sol | high | 98.7 | $0.765 | 11.4 |
| GPT-6 Astra | low | 98.4 | $1.53 | 8.6 |
| GPT 5.6 Sol | xhigh | 98.2 | $4.09 | 24.4 |
| GPT-6 Sol | max | 97.4 | $1.27 | 16.4 |
| GPT 5.5 | xhigh | 96.9 | $4.76 | 22.5 |
| GLM 5.3 (v3) | high | 96.7 | $1.61 | 33.8 |
| GPT 5.6 Sol | high | 96.4 | $3.31 | 19.9 |
| GLM 5.3 Flash (v3) | max | 95.9 | $0.298 | 49.2 |
| GPT-6 Sol | ultra | 95.9 | $2.52 | 16.1 |
| GPT 5.6 Sol | medium | 95.5 | $2.26 | 16.0 |
| Grok 4.6 | medium | 95.1 | $2.25 | 23.9 |
| Hy4 Preview | low | 95.1 | $0.651 | 35.9 |
| Claude Opus 5 | xhigh | 94.2 | $6.36 | 19.0 |
| K3 | high | 94.0 | $1.92 | 33.4 |
| Gemini 3.7 Flash | high | 94.0 | $2.00 | 15.1 |
| GLM 5.3 (v3) | max | 93.9 | $1.86 | 35.5 |
| GLM 5.3 Flash (v3) | high | 92.3 | $0.325 | 48.7 |
| GPT-6 Sol | xhigh | 92.3 | $0.941 | 13.0 |
| Gemini 3.8 Flash | medium | 92.1 | $2.13 | 16.2 |
| Gemini 3.8 Flash | high | 90.0 | $2.64 | 22.9 |
| GPT 5.5 | high | 89.7 | $3.12 | 16.2 |
| Gemini 3.7 Flash | medium | 89.5 | $1.69 | 14.4 |
| DeepSeek V4.1 Flash | high | 89.4 | $0.109 | 18.4 |
| K3 | low | 88.0 | $0.931 | 22.8 |
| GPT 5.6 Terra | xhigh | 88.0 | $1.53 | 19.3 |
| DeepSeek V4.1 Flash | max | 87.5 | $0.146 | 18.9 |
| Claude Opus 5 | max | 87.5 | $13 | 32.1 |
| GPT-6 Sol | medium | 86.3 | $0.541 | 9.2 |
| GPT 5.6 Luna | xhigh | 86.2 | $0.256 | 24.6 |
| DeepSeek V4 Flash | max | 85.7 | $0.211 | 31.3 |
| GPT 5.6 Sol | low | 81.3 | $1.16 | 10.4 |
| GLM 5.3 (v3) | low | 80.3 | $0.622 | 15.0 |
| GPT-6 Sol | low | 80.3 | $0.317 | 5.8 |
| GPT-6 Luna | max | 80.0 | $0.067 | 22.5 |
| Grok 4.6 | low | 78.6 | $0.989 | 11.9 |
| GPT 5.6 Terra | high | 77.2 | $0.898 | 13.6 |
| GPT 5.6 Luna | high | 76.3 | $0.162 | 17.3 |
| Gemini 3.7 Flash | low | 75.6 | $1.67 | 14.0 |
| Claude Sonnet 5 | medium | 75.0 | $1.93 | 9.8 |
| Claude Opus 5 | high | 75.0 | $5.73 | 17.3 |
| Hy4 Preview | max | 74.2 | $0.628 | 58.8 |
| GLM 5.3 Flash (v3) | low | 70.5 | $0.094 | 24.4 |
| Gemini 3.8 Flash | low | 67.5 | $3.09 | 17.4 |
| GPT-6 Luna | xhigh | 65.7 | $0.067 | 23.7 |
| DeepSeek V4 Flash | high | 65.6 | $0.187 | 30.4 |
| Kimi K2.8 Preview | low | 64.0 | — | 18.3 |
| Claude Sonnet 5 | low | 60.9 | $0.895 | 13.9 |
| GPT 5.6 Terra | medium | 59.8 | $0.547 | 9.4 |
| GPT 5.6 Terra | low | 50.9 | $0.439 | 8.0 |
| Claude Sonnet 5 | xhigh | 42.9 | $5.45 | 25.0 |
| GPT-6 Luna | high | 41.9 | $0.023 | 8.0 |
| GPT 5.6 Luna | medium | 32.6 | $0.066 | 10.7 |
| GPT-6 Luna | medium | 22.4 | $0.011 | 6.6 |
| GPT-6 Luna | low | 8.5 | $0.029 | 12.0 |
| GPT 5.6 Luna | low | 7.1 | $0.026 | 7.1 |
| Claude Sonnet 5 | max | 0.0 | $8.19 | 39.6 |
累计记录 50530 次运行。 2026年10月1日
Codex 雷达实测 · dsh
| 模型 | 推理档位 | 雷达 IQ | 平均费用(美元) | 平均用时(分钟) |
|---|---|---|---|---|
| Dsh Deepseek V4.1 Flash | high | 101.2 | $0.185 | 45.7 |
| Dsh Deepseek V4.1 Flash | max | 99.3 | $0.220 | 36.6 |
| Dsh Deepseek V4 Flash Vision Exp | max | 97.8 | $0.259 | 30.1 |
| Dsh Deepseek V4 Flash | max | 80.2 | $0.169 | 31.4 |
| Dsh Deepseek V4 Flash | high | 75.0 | $0.154 | 26.5 |
| Dsh Deepseek V4 Flash Vision Exp | high | 68.8 | $0.154 | 20.0 |
累计记录 50530 次运行。 2026年10月1日
社区评分
| 模型 | 平均分 / 10 | 评分次数 |
|---|---|---|
| GPT-6 Astra high | 10.0 | 1 |
| GPT-5.6 Sol medium | 10.0 | 1 |
| GPT-5.6 Terra ultra | 10.0 | 1 |
| GPT-5.6 Luna low | 10.0 | 1 |
| GPT-6 Astra xhigh | 9.0 | 2 |
| GPT-6 Astra ultra | 8.7 | 3 |
| GPT-6.1 Sol max | 8.6 | 7 |
| GPT-6.1 Sol high | 8.2 | 9 |
| GPT-6 Astra medium | 8.0 | 1 |
| GPT-5.6 Sol xhigh | 8.0 | 1 |
| GPT-5.6 Luna max | 8.0 | 1 |
| GPT-6.1 Sol xhigh | 8.0 | 5 |
| GPT-6.1 Sol medium | 8.0 | 1 |
| GPT-6 Astra max | 7.0 | 2 |
| GPT-6.1 Sol low | 6.0 | 2 |
| GPT-6 Sol xhigh | 4.0 | 1 |
2026年10月18日
Terminal-Bench 4.0 Resolution Rate % · 2026年9月21日
| 模型 | 运行环境 | 推理档位 | 分数 | 每题成本 |
|---|---|---|---|---|
| Claude Opus 5.5 | Claude Code | — | 66.4 | — |
| GPT-6 Astra | Codex | max | 58.2 | — |
| Claude Fable 5.1 | Claude Code | max | 57.9 | — |
| GPT-6 Astra | Codex | xhigh | 57.9 | — |
| GPT-6 Astra | Codex | high | 57.9 | — |
| Claude Fable 5.1 | Claude Code | xhigh | 57.9 | — |
| Claude Fable 5.1 | Claude Code | high | 54.5 | — |
| GPT-6 Astra | Codex | medium | 54.2 | — |
| Claude Fable 5.1 | Claude Code | medium | 53.9 | — |
| Claude Opus 5 | Claude Code | xhigh | 53.9 | — |
| Claude Opus 5 | Claude Code | max | 51.8 | — |
| GPT-6 Astra | Codex | low | 50.6 | — |
| Claude Opus 5 | Claude Code | high | 50.3 | — |
| Claude Opus 5 | Claude Code | medium | 44.9 | — |
| Claude Fable 5 | Claude Code | max | 44.5 | — |
| Claude Fable 5.1 | Claude Code | low | 43.3 | — |
| GLM 5.3 (v3) | Claude Code | max | 41.8 | — |
| Grok 4.7 | Grok Build | xhigh | 37.6 | — |
| GPT 5.6 Sol | Codex | max | 37.3 | — |
| Claude Opus 5 | Claude Code | low | 34.9 | — |
| DeepSeek V4.1 Flash | DeepSeek Harness Minimal | max | 31.2 | — |
| SWE-2 | Devin CLI | — | 27.3 | — |
| Claude Opus 4.8 | Claude Code | max | 23.6 | — |
| GPT 5.6 Terra | Codex | max | 21.5 | — |
| Grok 4.6 | Grok Build | high | 20.3 | — |
Artificial Analysis Coding Agent Index v1.5 Coding Agent Index · 2026年10月1日
| 模型 | 运行环境 | 推理档位 | 分数 | 每题成本 |
|---|---|---|---|---|
| Claude Sonnet 5.5 | Claude Code | max | 68.4 | $14 |
| Claude Opus 5.5 | Claude Code | max | 66.0 | $13 |
| GPT-6.1 Sol | Codex | xhigh | 62.9 | $1.04 |
| Claude Sonnet 5.5 | Claude Code | xhigh | 62.9 | $3.33 |
| Claude Fable 5.1 | Claude Code | max | 62.2 | $12 |
| Claude Fable 5.1 | Devin Fusion CLI | xhigh | 61.7 | $7.90 |
| GPT-6 Astra | Codex | max | 61.6 | $7.47 |
| GPT-6.1 Sol | Codex | medium | 61.4 | $0.705 |
| GPT-6.1 Sol | Codex | high | 60.1 | $0.889 |
| GPT-6.1 Sol | Codex | max | 60.1 | $1.55 |
| Claude Opus 5 | Claude Code | max | 59.7 | $11 |
| GPT-6 Astra | Devin Fusion CLI | xhigh | 58.9 | $4.54 |
| GPT-6.1 Sol | Codex | low | 57.2 | $0.499 |
| GPT-6 Sol | Codex | max | 56.7 | $2.99 |
| Grok 4.7 | Grok Build | xhigh | 56.3 | $8.82 |
| Claude Sonnet 5.5 | Claude Code | high | 55.0 | $1.24 |
| GPT 5.6 Sol | Codex | max | 54.6 | $6.35 |
| Muse Spark 1.3 | Muse Code | max | 54.3 | $3.98 |
| GLM 5.3 (v3) | Opencode | max | 53.6 | $4.24 |
| Kimi K3 (v1) | Kimi Code CLI | — | 51.9 | $5.05 |
| Muse Spark 1.3 | Muse Code 1.0.2 RC | xhigh | 48.3 | $3.47 |
| Grok 4.6 | Grok Build | xhigh | 47.0 | $3.57 |
| Claude Sonnet 5.5 | Claude Code | medium | 45.9 | $0.619 |
| Qwen3.8 Max | Claude Code | max | 43.3 | $3.48 |
| GPT 5.6 Luna | Codex | max | 43.2 | $0.438 |
DeepSWE v1.1 Pass@1 % · 2026年9月3日
| 模型 | 运行环境 | 推理档位 | 分数 | 每题成本 |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | mini-SWE | max | 74.2 | — |
| GPT-6 Astra | mini-swe-agent | xhigh | 74.1 | $6.52 |
| Gemini 3.8 Flash | mini-swe-agent | high | 73.8 | $2.36 |
| Claude Opus 5 | mini-swe-agent | max | 73.6 | $12 |
| GPT-6 Astra | mini-swe-agent | high | 73.2 | $5.72 |
| GPT-6 Astra | mini-swe-agent | max | 73.2 | $12 |
| Claude Opus 5 | mini-swe-agent | xhigh | 73.2 | $9.07 |
| Claude Opus 5 | mini-swe-agent | high | 72.8 | $6.08 |
| GPT-6 Astra | mini-swe-agent | medium | 72.8 | $4.38 |
| GPT 5.6 Sol | mini-swe-agent | max | 72.7 | $8.39 |
| MiMo V2.6 Pro | mini-swe-agent | — | 71.9 | — |
| Gemini 3.8 Flash | mini-swe-agent | medium | 71.0 | $1.97 |
| Grok 4.7 | mini-swe-agent | high | 71.0 | — |
| GPT 5.6 Sol | mini-swe-agent | xhigh | 70.7 | $4.70 |
| Claude Fable 5 | mini-swe-agent | xhigh | 69.9 | $13 |
| Claude Fable 5 | mini-swe-agent | max | 69.7 | $22 |
| GPT 5.6 Terra | mini-swe-agent | max | 69.6 | $4.95 |
| GPT 5.6 Sol | mini-swe-agent | high | 69.4 | $3.47 |
| GLM 5.3 (v3) | mini-swe-agent | max | 69.0 | $3.99 |
| Claude Opus 5 | mini-swe-agent | medium | 68.9 | $3.29 |
| Claude Fable 5 | mini-swe-agent | high | 68.6 | $9.18 |
| Kimi K3 (v1) | mini-swe-agent | max | 68.5 | $4.65 |
| MiMo V2.6 Flash | mini-swe-agent | — | 67.9 | — |
| Grok 4.6 | mini-swe-agent | medium | 67.5 | $3.45 |
| GPT 5.6 Luna | mini-swe-agent | max | 67.2 | $3.03 |