Eval Connections — Guess Accuracy
Games · 2025-07-31
Eval Connections Guess Accuracy is the percentage of submitted four-word guesses that match one of the target groups across the same 20-game canonical battery. The denominator varies with the number of guesses each model makes.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 100.0 |
| Sonnet 4.6 | 100.0 |
| Gemini 3.1 Pro Preview | 100.0 |
| GPT-5.2 Pro | 100.0 |
| GPT-5.6 Sol Pro | 100.0 |
| OpenRouter Fusion | 100.0 |
| Qwen3.7-Plus | 100.0 |
| Sonnet 5 | 98.8 |
| Gemini 3 Pro Preview | 98.8 |
| Opus 4.7 | 97.6 |
| GPT-5.6 Terra | 97.6 |
| GPT-5.6 Terra Pro | 97.6 |
| Grok 4 | 97.6 |
| Grok 4.20 | 97.6 |
| o3 | 97.6 |
| Opus 4.6 | 97.1 |
| GLM-5.1 | 96.4 |
| GPT-5.6 Sol | 96.4 |
| DeepSeek-V4-Pro | 95.2 |
| Gemini 3.5 Flash | 95.2 |
| Gemini 3 Flash Preview | 95.2 |
| GPT-5.3-Codex | 95.2 |
| Grok 4 Fast | 94.1 |
| GLM-5.2 | 93.0 |
| Opus 4.5 | 92.9 |
| GPT-5.3 Instant | 90.7 |
| Kimi K2.5 | 89.9 |
| Qwen3.6 Plus (2026-04-02) | 89.9 |
| Grok 4.3 | 89.7 |
| Qwen3 Max Thinking (2026-01-23) | 89.7 |
| MiniMax M3 | 88.4 |
| GLM-4.7 | 87.5 |
| DeepSeek-V4-Flash | 86.5 |
| GLM-5 | 86.5 |
| GPT-5.6 Luna Pro | 85.7 |
| Step 3.7 Flash | 85.6 |
| Gemini 2.5 Pro | 84.3 |
| Kimi K2.6 | 83.7 |
| GPT-5.6 Luna | 83.1 |
| Kimi K2.7 Code | 82.8 |