Skip to content
localmodel.run

Leaderboard · LMArena

Best local LLMs by human preference

LMArena ranks models by Elo from blind, head-to-head human votes. It captures overall chat quality and instruction-following as people actually perceive it, which is why it is the most-cited general leaderboard.

16 models · ranked by LMArena runs on
  1. 1 1383 Qwen3 30B-A3B 30.5B MoE
  2. 2 1366 Gemma 3 27B 27B
  3. 3 1357 Mistral Small 3 24B 24B
  4. 4 1347 Qwen3 32B 32B
  5. 5 1342 Gemma 3 12B 12B
  6. 6 1318 Llama 3.3 70B 70B
  7. 7 1303 Gemma 3 4B 4B
  8. 8 1303 Qwen2.5 72B 72B
  9. 9 1289 Gemma 2 27B 27B
  10. 10 1266 Gemma 2 9B 9B
  11. 11 1256 Phi-4 14B 14B
  12. 12 1211 Llama 3.1 8B 8B
  13. 13 1166 Llama 3.2 3B 3B
  14. 14 1149 Mistral 7B 7B
  15. 15 1114 SmolLM2 1.7B 1.7B
  16. 16 1110 Llama 3.2 1B 1B

Score is LMArena (human-preference elo from lmarena's blind head-to-head votes.), sourced from lmarena.ai/leaderboard. The "runs on" column is the lightest tracked device that runs the model at Q4_K_M; tap a model for the full hardware list. Coverage is partial: only models we verified against the canonical board appear.

Other boards

Best local LLMs for coding Percent of 225 Exercism exercises solved across six languages. Best local LLMs for tool use and agents Overall Acc on the Berkeley Function-Calling Leaderboard.

Or pick by what you need with use-case picks, or browse the full model list.

FAQ

What is the best local model for general chat?

Qwen3 30B-A3B tops this board at 1383 on LMArena. It needs ~20.7 GB at Q4_K_M, so the lightest hardware that runs it is Apple M5 (32GB).

How is this chat (elo) leaderboard scored?

LMArena ranks models by Elo from blind, head-to-head human votes. It captures overall chat quality and instruction-following as people actually perceive it, which is why it is the most-cited general leaderboard. We only list a model after verifying its score against the canonical leaderboard for that exact model, so coverage is partial by design. Source: lmarena.ai/leaderboard.

Can I run these models locally?

Every row shows the lightest tracked device that runs the model at Q4_K_M and links to the full hardware list, so you see the ranking and your hardware fit together.

Benchmark scores are third-party, sourced from lmarena.ai/leaderboard, and reflect the full-precision model, not a specific quant. Memory and fit are computed at Q4_K_M, validated 2026-08-03. See methodology.