Artificial Analysis Coding Agent Index v2.0
Indexes
Artificial Analysis Coding Agent Index v2.0 scores coding-agent and model pairs, such as Claude Code or Codex with a given model and effort, as the equal-weight average of mean scores over three attempts per task on FrontierCode v1.1 (100 tasks), Terminal-Bench 4.0 (66 tasks) and SWE-Atlas-QnA-AA (124 tasks, grading adjusted by Artificial Analysis); attempts flagged for reward hacking score zero. FrontierCode v1.1 replaces v1.5's DeepSWE v1.1, so v2.0 scores are not comparable with v1.5. Scores belong to the agent harness and model together. It is separate from the model-level Artificial Analysis Coding Index.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 61.7 |
| Claude Sonnet 5.5 | 59.3 |
| Muse Wonder Full | 58.9 |
| Claude Fable 5.1 | 57.6 |
| GPT-6 Astra | 54.7 |
| GPT-6.1 Sol | 54.1 |
| Gemini 4 Argon | 53.2 |
| Muse Wonder | 52.0 |
| GLM-5.3 | 48.9 |
| DeepSeek V4.1 Flash | 47.3 |
| Grok 4.7 | 46.6 |
| DeepSeek V4.1 Flash 1017 | 46.3 |
| Muse Spark 1.3 | 45.7 |
| Step 5 Preview | 43.1 |
| Kimi K3 | 41.4 |
| Claude Haiku 5.5 | 37.3 |
| Gemini 3.8 Flash | 33.3 |
| GPT-6 Luna | 31.1 |
| Muse Spark Open | 26.3 |
| Beam 1.0 501B A23B | 20.5 |