FMG-Bench - Raw Model: leaderboard
Metric: Judge-panel score (0-100; no benchmark guidance). Source: github.com. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 92.33 |
| 2 | GPT-5.4 | 91.79 |
| 3 | Kimi K2.6 | 90.61 |
| 4 | Grok 4.20 | 90.37 |
| 5 | Qwen 3.6 Plus | 89.93 |
| 6 | DeepSeek V4 Pro | 89.13 |
| 7 | GLM-5.1 | 88.89 |
| 8 | Gemini 3.1 Pro (Preview) | 87.72 |
| 9 | Nemotron 3 Super | 87.18 |
| 10 | MiMo-V2.5-Pro | 86.94 |
| 11 | MiniMax-M2.7 | 84.61 |
| 12 | Mistral Large 3 | 84.25 |
| 13 | Seed 2.0 Lite | 83.13 |
| 14 | Llama 4 Maverick | 73.51 |
Interactive version: theaggregate.ai/benchmark?slug=fmg-bench-raw-model · How It Works · Data refreshed daily, snapshot 2026-09-19.