Eval Connections
Games · 2025-07-31
Eval Connections measures the percentage of 20 fixed word-grouping games a model solves. Each game presents 16 shuffled words forming four groups of four; the model receives multi-turn correctness feedback and must recover all four groups before making four mistakes.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 100.0 |
| Opus 4.7 | 100.0 |
| Sonnet 4.6 | 100.0 |
| Sonnet 5 | 100.0 |
| DeepSeek-V4-Pro | 100.0 |
| Fugu Ultra v1.0 | 100.0 |
| Gemini 3.1 Pro Preview | 100.0 |
| Gemini 3.5 Flash | 100.0 |
| Gemini 3 Flash Preview | 100.0 |
| Gemini 3 Pro Preview | 100.0 |
| GLM-5.1 | 100.0 |
| GLM-5.2 | 100.0 |
| GPT-5.2 Pro | 100.0 |
| GPT-5.3-Codex | 100.0 |
| GPT-5.4 | 100.0 |
| GPT-5.6 Sol | 100.0 |
| GPT-5.6 Sol Pro | 100.0 |
| GPT-5.6 Terra | 100.0 |
| GPT-5.6 Terra Pro | 100.0 |
| Grok 4 | 100.0 |
| Grok 4.20 | 100.0 |
| Grok 4 Fast | 100.0 |
| Kimi K2.5 | 100.0 |
| o3 | 100.0 |
| OpenRouter Fusion | 100.0 |
| Qwen3.6 Plus (2026-04-02) | 100.0 |
| Qwen3.7-Plus | 100.0 |
| Opus 4.5 | 95.0 |
| GLM-4.7 | 95.0 |
| GLM-5 | 95.0 |
| GPT-5.3 Instant | 95.0 |
| GPT-5.6 Luna Pro | 95.0 |
| Grok 4.3 | 95.0 |
| Grok 4.5 | 95.0 |
| Kimi K2.6 | 95.0 |
| Kimi K2.7 Code | 95.0 |
| Qwen3 Max Thinking (2026-01-23) | 95.0 |
| Step 3.7 Flash | 95.0 |
| DeepSeek-V4-Flash | 90.0 |
| Gemini 2.5 Pro | 90.0 |