# general
gemma2:9b
gemma2:9bFast general queries, lookups & summaries. ~12 tok/s.
qwen2.5:14b
qwen2.5:14bStrong reasoning, structured output, multi-step logic. ~7 tok/s.
mistral-nemo:12b
mistral-nemo:12bTechnical & infrastructure tasks, sysadmin queries. ~9 tok/s.
deepseek-coder-v2:16b
deepseek-coder-v2:16bCode generation, review & debugging. ~7 tok/s.
llama3.3:70b
llama3.3:70bHighest quality answers. Slowest (~2 tok/s, ~8 min cold load).
llama3.2:3b
llama3.2:3bUltra-fast simple lookups. ~25 tok/s.