8.1 Global

deepseek-r1:32b

Judge: gemma4:31b · 144/160 tests · 1 h 13 min · 10.9 tok/s

32.8B · Q4_K_M · 18.5 GB · 131K ctx

Thinking

Category breakdown

surprise 10.0

long-context 9.9

agentic 9.4

instruction 8.7

code 8.4

reasoning 8.2

math 8.0

multilingual 7.9

frontend 7.8

safety 7.6

organization 7.5

roleplay 7.4

writing 6.1