EnigmaForge - Fact F1: leaderboard
EnigmaForge's world-reconstruction score: F1 of the world facts the model recovers from the story, over all 600 instances with missing answers counted as zero; higher is better.
Metric: World-fact recovery F1 (%) over 600 instances, all items. Source: robottwo.github.io. Saturation forecast: Estimated already saturated. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6.1 Sol (Low) | 98.5 | #10 (GPT-6.1 Sol) |
| 2 | Gemini 4 Argon (Low) | 98.3 | #11 (Gemini 4 Argon) |
| 3 | GPT-6.1 Terra (Low) | 97.8 | #20 (GPT-6.1 Terra) |
| 4 | Claude Fable 5.2 (Low) | 97.2 | #1 (Claude Fable 5.2) |
| 5 | GPT-6 Sol (Low) | 96.6 | #25 (GPT-6 Sol) |
| 6 | Claude Opus 5.5 (Low) | 91.9 | #5 (Claude Opus 5.5) |
| 7 | Gemini 3.7 Flash (Low) | 91.3 | #32 (Gemini 3.7 Flash) |
| 8 | GPT-5.6 Sol (Low) | 91 | #30 (GPT-5.6 Sol) |
| 9 | Gemini 4 Carbon (Low) | 91 | #2 (Gemini 4 Carbon) |
| 10 | GPT-6 Astra (Low) | 90.6 | #6 (GPT-6 Astra) |
| 11 | Grok 4.6 (Low) | 89.5 | #45 (Grok 4.6) |
| 12 | Muse Wonder (Low) | 87.9 | #28 (Muse Wonder) |
| 13 | Gemini 3.8 Flash (Low) | 86.2 | #36 (Gemini 3.8 Flash) |
| 14 | Qwen 3.8 27B (Low) | 83.6 | #204 (Qwen 3.8 27B) |
| 15 | GPT-5.6 Terra (Low) | 81.8 | #62 (GPT-5.6 Terra) |
Interactive version: theaggregate.ai/benchmark?slug=enigmaforge-fact-f1 · How It Works · Data refreshed daily, snapshot 2026-11-13.