Eval Connections One-Shot Group Accuracy
Games · 2026-07-22
Percentage of 80 canonical word groups recovered across 20 Eval Connections one-shot submissions. No correctness feedback; current structural-format repair allowance is two turns. Correlated component of the one-shot points metric.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 100.0 |
| Muse Spark 1.1 | 100.0 |
| Gemini 3.8 Flash | 100.0 |
| Claude Opus 5.5 | 100.0 |
| Gemini 3.1 Pro Preview | 100.0 |
| GPT-5.6 Terra | 100.0 |
| Gemini 3 Flash Preview | 100.0 |
| GPT-5.6 Sol Pro | 100.0 |
| Fugu Ultra v1.0 | 100.0 |
| Grok 4.20 | 100.0 |
| Gemini 3.7 Flash | 100.0 |
| DeepSeek V4 Pro 0813 | 100.0 |
| GLM-5.3 Flash | 100.0 |
| Claude Opus 5 | 100.0 |
| Grok 4.6 | 97.5 |
| GPT-6 Sol | 97.5 |
| Claude Fable 5 | 97.5 |
| Grok 4.7 | 97.5 |
| GPT-5.6 Terra Pro | 97.5 |
| Muse Spark 1.3 | 96.3 |
| Claude Fable 5.1 | 96.3 |
| Muse Spark 1.2 | 96.3 |
| DeepSeek-V4-Flash-0731 | 95.0 |
| Qwen3.8 Flash | 95.0 |
| Gemini 3.6 Flash | 95.0 |
| Opus 4.7 | 95.0 |
| Kimi K3 | 92.5 |
| Grok 4.5 | 92.5 |
| Opus 4.5 | 92.5 |
| GPT-5.6 Luna Pro | 91.3 |
| GPT-5.6 Sol | 90.0 |
| Sonnet 5 | 88.8 |
| GLM 5V Turbo | 87.5 |
| DeepSeek V4.1 Flash | 87.5 |
| Gemini 3.5 Flash-Lite | 86.3 |
| Gemini 3.5 Flash | 85.0 |
| GLM-5.3 | 85.0 |
| Inkling | 83.8 |
| MiMo-V2.6-Flash | 82.5 |
| GPT-6 Luna | 81.3 |