FrontierMath Tier 4 (v2) - Private
Math · 2026-06-12
The Private split of FrontierMath Tier 4 (v2). This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6.1 Sol | 100.0 |
| GPT-6 Astra | 97.6 |
| Claude Opus 5.5 | 95.0 |
| Claude Fable 5 | 90.2 |
| GPT-6 Sol | 90.0 |
| Claude Fable 5.1 | 87.8 |
| GPT-5.6 Sol | 83.0 |
| Claude Sonnet 5.5 | 80.5 |
| GPT-5.6 Sol Pro | 80.5 |
| Gemini 4 Argon | 78.0 |
| GPT-5.5 Pro | 78.0 |
| Claude Opus 5 | 73.2 |
| GPT-5.5 | 72.5 |
| GPT-5.6 Terra | 70.7 |
| Muse Wonder | 70.7 |
| GPT-5.6 Luna | 61.0 |
| GPT-5.4 Pro | 58.5 |
| Opus 4.8 | 56.1 |
| GPT-6 Luna | 56.1 |
| GPT-5.4 | 49.0 |
| Claude Haiku 5.5 | 46.3 |
| DeepSeek V4.1 Flash 1017 | 46.3 |
| Muse Spark 1.3 | 46.3 |
| Qwen3.8-Max | 46.3 |
| GPT-5.2 Pro | 46.0 |
| Kimi K3 | 39.0 |
| Gemini 3.7 Flash | 36.6 |
| Qwen3.7-Max (source unspecified) | 34.1 |
| Qwen3.8 Max (0902) | 34.1 |
| Opus 4.7 | 31.7 |
| Grok 4.6 | 31.7 |
| Muse Spark Open | 31.7 |
| GPT-5.2 | 31.7 |
| Sonnet 5 | 29.3 |
| GLM-5.2 | 29.3 |
| GLM-5.3 | 29.3 |
| Opus 4.6 | 26.8 |
| DeepSeek V4.1 Flash | 26.8 |
| DeepSeek V4 Pro 0813 | 26.8 |
| Gemini 3.1 Pro Preview | 26.8 |