Calibration
Vibe Code Bench → FrontierCode 1.1 Main
FrontierCode 1.1 Main is estimated from Vibe Code Bench with a inverse Michaelis–Menten curve fitted on 5 models measured on both: y = 0.74004·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.85, cross-validated error 4.3 pp. It is used for 20 estimates.
| Estimated model | Vibe Code Bench | FrontierCode 1.1 Main | Source |
|---|---|---|---|
| Claude Haiku 4.5 Thinking | 11.4% | 4.5% | estimated ± 4.3 pp, low confidence |
| Claude Opus 4.5 Thinking | 20.6% | 8.5% | estimated ± 4.3 pp, low confidence |
| Claude Opus 4.6 (Adaptive) | 53.5% | 27.0% | estimated ± 4.3 pp, medium confidence |
| Claude Sonnet 4.5 Thinking | 22.6% | 9.4% | estimated ± 4.3 pp, low confidence |
| DeepSeek V3.2 (Thinking) | 5.1% | 1.9% | estimated ± 4.3 pp, low confidence |
| Gemini 3.1 Flash-Lite | 0.0% | 0.0% | estimated ± 4.3 pp, low confidence |
| Gemini 3 Flash | 20.2% | 8.3% | estimated ± 4.3 pp, low confidence |
| Gemini 3 Pro | 14.3% | 5.7% | estimated ± 4.3 pp, low confidence |
| GLM-4.6 | 3.1% | 1.2% | estimated ± 4.3 pp, low confidence |
| GLM-5 (Reasoning) | 23.4% | 9.8% | estimated ± 4.3 pp, low confidence |
| GPT-5.1-Codex | 13.1% | 5.2% | estimated ± 4.3 pp, low confidence |
| GPT-5.1-Codex-Max | 22.2% | 9.2% | estimated ± 4.3 pp, low confidence |
| GPT-5.2-Codex | 37.9% | 17.3% | estimated ± 4.3 pp, low confidence |
| GPT-5 mini | 14.2% | 5.6% | estimated ± 4.3 pp, low confidence |
| Grok 4.1 Fast (Reasoning) | 1.2% | 0.4% | estimated ± 4.3 pp, low confidence |
| Grok 4 Fast (Reasoning) | 0.0% | 0.0% | estimated ± 4.3 pp, low confidence |
| MiniMax M2.5 | 14.9% | 5.9% | estimated ± 4.3 pp, low confidence |
| Mistral Large 4 | 78.4% | 47.7% | estimated ± 4.3 pp, low confidence |
| Qwen3.5 Plus | 15.7% | 6.3% | estimated ± 4.3 pp, low confidence |
| Qwen3 Max | 3.5% | 1.3% | estimated ± 4.3 pp, low confidence |