← Dashboard

LLM Quality Scores

lm-eval quick panel · gsm8k = exact_match (flexible) · ifeval = prompt_level_strict_acc

Modelgsm8kifeval-strictDate
Qwen-AgentWorld-35B-A3B-GGUF:latest0.91740.74682026-07-17
qwen3.6:ctx131k0.90520.83922026-07-18
qwen3.6:35b-a3b-mtp-q4_K_M0.89840.81332026-07-17
qwen3-coder:ctx64k0.89310.80592026-07-18
ornith:ctx131k0.87950.83732026-07-18
gemma-26b:ctx131k0.81960.86882026-07-18