Arena

Race up to four models on the same prompt to compare speed and answer quality in LM Mini Arena.

Answer first

Arena runs the same prompt across multiple models at once so you can compare tokens/sec, time-to-first-token, and the written answers — then pick a default for chat or Personas.

Open Arena

What you can race

Add contestants from this device (including Apple Intelligence), LM Studio, Ollama, JAN AI, oMLX, Unsloth, or a cloud provider. Arena is free: race up to four models on the built-in prompts or Your prompt, compare the answers side by side, and keep a history of your races. Cloud providers that are themselves Pro stay locked, the same as in chat.

Suggested lanes: this device vs desktop, “Race top 2”, Quick vs Full suite — pick what you have loaded.

How a race works

  1. Pick contestants and a prompt (or use the built-in discovery prompts)
  2. Start — streams run live with speed metrics
  3. Read answers, note TTFT / tok/s
  4. Optionally apply that model as your everyday default or Persona binding

Anonymous benchmarks

Arena can share anonymous speed results — model, device class, and speed numbers, never your prompts or answers — to build a community leaderboard. You choose once before your first race, and can turn it off anytime under Settings → Advanced → Share Arena speed results.