← Dashboard

LLM Quality Scores

lm-eval aggregate scores (· run time where known) · reasoning = bbh_cot_zeroshot (flexible) · knowledge = mmlu_pro · math = gsm8k · instruct = ifeval-strict. Run reasoning in thinknothink reasoning is a floor, not capability.

ModelModeReasoningKnowledgeMathInstructDate
Qwen-AgentWorld-35B-A3B-GGUF:latestnothink0.9170.7472026-07-17
agentworld:ctx131knothink0.5930.829 · 12m0.9160.7522026-07-27
agentworld:ctx131kthink0.896 · 88m0.893 · 17m2026-07-22
apodex:ctx131knothink0.447 · 21m0.763 · 19m2026-08-31
apodex:ctx131kthink0.837 · 154m0.789 · 118m2026-08-31
gemma-12b:ctx131knothink0.407 · 5m0.714 · 13m2026-07-27
gemma-26b:ctx131knothink0.430 · 32m0.814 · 10m0.820 · 24m0.8692026-07-27
glm-flash-xl:ctx64knothink0.617 · 21m2026-08-08
glm-flash:ctx64knothink0.614 · 20m0.793 · 116m2026-08-08
glm-flash:ctx64kthink0.792 · 237m2026-08-04
gpt-oss-20b:ctx131knothink0.7780.354 · 136m0.8472026-08-07
gpt-oss:ctx64knothink0.8520.500 · 16m0.3790.7632026-07-27
gpt-oss:ctx64kthink0.830 · 63m0.571 · 14m2026-07-22
laguna:ctx131knothink0.399 · 12m0.591 · 149m0.793 · 24m2026-08-13
laguna:ctx131kthink0.792 · 2029m2026-08-01
muse-glimmer:bench8knothink0.7532026-08-15
nemotron-omni:ctx131knothink0.477 · 28m0.7970.8082026-08-08
nemotron-omni:ctx131kthink0.733 · 29m2026-07-22
nemotron-q5:ctx131knothink0.529 · 28m2026-08-08
ornith15:ctx131knothink0.760 · 12m0.763 · 48m2026-08-22
ornith15:ctx131kthink0.8822026-08-21
ornith:ctx131knothink0.743 · 9m0.8790.8372026-07-27
ornith:ctx131kthink0.889 · 33m2026-07-21
qwen3-coder:ctx64knothink0.593 · 3m0.671 · 4m0.8930.8062026-07-27
qwen3.6-xl:ctx131knothink0.779 · 52m2026-08-09
qwen3.6:35b-a3b-mtp-q4_K_Mnothink0.8980.8132026-07-17
qwen3.6:bench32kthink0.815 · 166m2026-08-30
qwen3.6:ctx131knothink0.351 · 19m0.792 · 53m0.9050.8262026-08-31
qwen3.6:ctx131kthink0.926 · 137m0.844 · 168m2026-08-31
qwen3.8ud:bench32kthink0.8850.825 · 251m2026-08-29
qwen3.8ud:bench8knothink0.837 · 132m2026-08-22
qwen3.8ud:ctx131k-mtpnothink0.779 · 100m2026-08-22
qwen38-flash-nextnothink0.3960.8210.8522026-08-27
qwen38-flash-nextthink0.8852026-08-28