Calibration
Vibe Code Bench → AA Coding Index
AA Coding Index is estimated from Vibe Code Bench with a linear curve fitted on 16 models measured on both: y = 0.5460·x + 0.3864, R² = 0.73, cross-validated error 6.8 pp. It is used for 16 estimates.
| Estimated model | Vibe Code Bench | AA Coding Index | Source |
|---|---|---|---|
| Claude Opus 4.6 (Adaptive) | 53.5% | 67.8% | estimated ± 6.8 pp, medium confidence |
| GLM-5 (Reasoning) | 23.4% | 51.4% | estimated ± 6.8 pp, medium confidence |
| Claude Sonnet 4.5 Thinking | 22.6% | 51.0% | estimated ± 6.8 pp, medium confidence |
| GPT-5.1-Codex-Max | 22.2% | 50.7% | estimated ± 6.8 pp, medium confidence |
| Claude Opus 4.5 Thinking | 20.6% | 49.9% | estimated ± 6.8 pp, medium confidence |
| Qwen3.5 Plus | 15.7% | 47.2% | estimated ± 6.8 pp, medium confidence |
| MiniMax M2.5 | 14.9% | 46.7% | estimated ± 6.8 pp, medium confidence |
| Gemini 3 Pro | 14.3% | 46.4% | estimated ± 6.8 pp, medium confidence |
| GPT-5 mini | 14.2% | 46.4% | estimated ± 6.8 pp, medium confidence |
| GPT-5.1-Codex | 13.1% | 45.8% | estimated ± 6.8 pp, medium confidence |
| Claude Haiku 4.5 Thinking | 11.4% | 44.9% | estimated ± 6.8 pp, medium confidence |
| DeepSeek V3.2 (Thinking) | 5.1% | 41.4% | estimated ± 6.8 pp, medium confidence |
| Qwen3 Max | 3.5% | 40.5% | estimated ± 6.8 pp, medium confidence |
| GLM-4.6 | 3.1% | 40.3% | estimated ± 6.8 pp, medium confidence |
| Grok 4.1 Fast (Reasoning) | 1.2% | 39.3% | estimated ± 6.8 pp, medium confidence |
| Grok 4 Fast (Reasoning) | 0.0% | 38.6% | estimated ± 6.8 pp, low confidence |
Fitted on
| Model | Vibe Code Bench | AA Coding Index |
|---|---|---|
| DeepSeek V4 Pro 0813 | 49.9% | 68.8% |
| Gemini 2.5 Pro | 0.4% | 33.3% |
| Gemini 3.1 Pro | 32.0% | 68.8% |
| Gemini 3.5 Flash | 48.7% | 70.1% |
| GLM-5.1 | 31.5% | 55.8% |
| GPT-5.1 | 24.6% | 49.4% |
| GPT-5.4 | 67.4% | 71.1% |
| GPT-5.4 mini | 48.0% | 56.1% |
| GPT-5.4 nano | 26.1% | 56.1% |
| GPT-5.5 | 69.8% | 74.9% |
| GPT-5 (high) | 20.1% | 37.8% |
| Kimi K2.6 | 37.9% | 61.8% |
| Kimi K2.5 (Reasoning) | 17.5% | 46.8% |
| MiniMax M2.7 | 27.0% | 52.6% |
| Muse Spark | 19.7% | 58.6% |
| Qwen3.6 Plus | 25.6% | 54.5% |