MLCR-AA Completeness (Judged Responses)
Professional Work
MLCR-AA completeness among responses sent to the judge panel. It uses the conditional judged-response denominator and does not penalize excluded overlong or empty responses.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 89.0 |
| Claude Sonnet 5.5 | 88.4 |
| Claude Opus 5.5 | 88.4 |
| Claude Opus 5 | 86.1 |
| Claude Haiku 5.5 | 81.3 |
| Claude Fable 5 | 73.8 |
| Sonnet 5 | 68.2 |
| GLM-5.3 Flash | 63.8 |
| GLM-5.3 | 62.0 |
| Sonnet 4.6 | 57.9 |
| Opus 4.8 | 52.0 |
| Muse Spark 1.2 | 51.7 |
| Muse Spark 1.3 | 49.7 |
| Kimi K3 | 45.1 |
| MiMo-V2.6-Pro | 42.1 |
| Opus 4.7 | 40.3 |
| GPT-6 Astra | 36.8 |
| GPT-6.1 Sol | 36.2 |
| MiniMax M3 | 34.7 |
| Muse Glimmer 30B | 33.9 |
| GPT-5.6 Terra | 33.9 |
| GPT-5.6 Sol | 27.6 |
| DeepSeek V4.1 Flash | 27.0 |
| Gemini 3.8 Flash | 26.9 |
| DeepSeek-V4-Pro | 26.3 |
| Qwen3.8 27B | 25.6 |
| Inkling | 24.7 |
| DeepSeek V4 Pro 0813 | 23.4 |
| Gemini 3.5 Flash | 23.3 |
| Qwen3.8-Max | 23.2 |
| Qwen3.8 Max (0902) | 22.1 |
| GPT-5.6 Luna | 21.8 |
| Kimi K2.7 Code | 21.6 |
| Step 5 Preview | 21.3 |
| Gemini 3.7 Flash | 20.9 |
| Nemotron 3 Ultra 550B A55B | 20.5 |
| Gemini 3.6 Flash | 19.8 |
| GPT-5.5 | 19.5 |
| DeepSeek-V4-Flash | 19.4 |
| Grok 4.7 | 18.8 |