Calibration
Vibe Code Bench → SWE-bench Verified
SWE-bench Verified is estimated from Vibe Code Bench with a linear curve fitted on 11 models measured on both: y = 0.1886·x + 0.7165, R² = 0.61, cross-validated error 4.4 pp. It is used for 25 estimates.
| Estimated model | Vibe Code Bench | SWE-bench Verified | Source |
|---|---|---|---|
| Claude Haiku 4.5 Thinking | 11.4% | 73.8% | estimated ± 4.4 pp, high confidence |
| Claude Opus 4.5 Thinking | 20.6% | 75.5% | estimated ± 4.4 pp, high confidence |
| Claude Opus 4.6 (Adaptive) | 53.5% | 81.7% | estimated ± 4.4 pp, high confidence |
| Claude Sonnet 4.5 Thinking | 22.6% | 75.9% | estimated ± 4.4 pp, high confidence |
| DeepSeek V3.2 (Thinking) | 5.1% | 72.6% | estimated ± 4.4 pp, high confidence |
| Gemini 3.1 Flash-Lite | 0.0% | 71.7% | estimated ± 4.4 pp, medium confidence |
| Gemini 3.1 Pro | 32.0% | 77.7% | estimated ± 4.4 pp, high confidence |
| Gemini 3 Flash | 20.2% | 75.5% | estimated ± 4.4 pp, high confidence |
| Gemini 3 Pro | 14.3% | 74.4% | estimated ± 4.4 pp, high confidence |
| Gemini 4 Argon | 91.9% | 89.0% | estimated ± 4.4 pp, medium confidence |
| GLM-4.6 | 3.1% | 72.2% | estimated ± 4.4 pp, high confidence |
| GLM-5 (Reasoning) | 23.4% | 76.1% | estimated ± 4.4 pp, high confidence |
| GPT-5.1 | 24.6% | 76.3% | estimated ± 4.4 pp, high confidence |
| GPT-5.1-Codex | 13.1% | 74.1% | estimated ± 4.4 pp, high confidence |
| GPT-5.1-Codex-Max | 22.2% | 75.8% | estimated ± 4.4 pp, high confidence |
| GPT-5.2-Codex | 37.9% | 78.8% | estimated ± 4.4 pp, high confidence |
| GPT-5.4 nano | 26.1% | 76.6% | estimated ± 4.4 pp, high confidence |
| GPT-5 (high) | 20.1% | 75.4% | estimated ± 4.4 pp, high confidence |
| GPT-5 mini | 14.2% | 74.3% | estimated ± 4.4 pp, high confidence |
| Grok 4.1 Fast (Reasoning) | 1.2% | 71.9% | estimated ± 4.4 pp, high confidence |
| Grok 4 Fast (Reasoning) | 0.0% | 71.7% | estimated ± 4.4 pp, medium confidence |
| MiniMax M2.5 | 14.9% | 74.5% | estimated ± 4.4 pp, high confidence |
| Mistral Large 4 | 78.4% | 86.4% | estimated ± 4.4 pp, medium confidence |
| Qwen3.5 Plus | 15.7% | 74.6% | estimated ± 4.4 pp, high confidence |
| Qwen3 Max | 3.5% | 72.3% | estimated ± 4.4 pp, high confidence |