localLLMrig

Free testing · no sign-up

Test AI models side by side

Choose your LLMs, ask once and compare the answers. Pick the best one and see the PC, Mac or server that runs it.

The models are hosted by third-party providers who may log prompts. Don't enter personal or confidential info. Privacy notice.

Choose LLMs

Ranked by what it costs to run each model yourself: the price of the cheapest PC, Mac or server that holds it (Q4, 8,192 tokens). Estimates, not benchmarks. Free: tick 2 or more. 35 more with a pass.

Tick the free models to test

Under $1,0001 free · 9 with a pass

Small PC or Mac. Runs on the Mac mini M4 16GB, from $615.

  • Ministral 3 14BMistral · 14B
  • Qwen3 14BQwen · 14B
  • Gemma 3 12BGoogle · 12B
  • Qwen3.5 9BQwen · 9B
  • Llama 3.1 8B InstructMeta · 8.03B
  • Gemma 3 4BGoogle · 4B
  • Llama 3.2 3B InstructMeta · 3.21B
  • Ministral 3 3BMistral · 3B
  • Llama 3.2 1B InstructMeta · 1B

$1,000–2,0006 free · 8 with a pass

Desktop PC or Mac. From the Budget PC with RTX 3060 12GB ($1,519) up to the Mac mini M6 32GB ($1,713) for the largest.

  • GLM 4.7 FlashZ.ai · 30B, 3B active
  • Nemotron 3 Nano 30B A3BNVIDIA · 30B, 3.5B active
  • Qwen3 Coder 30B A3BQwen · 30B, 3B active
  • Gemma 3 27BGoogle · 27B
  • LFM2.5 24B A2BLiquid AI · 24B, 2.45B active
  • Mistral Small 3.2 24BMistral · 24B
  • gpt-oss-20bOpenAI · 21B, 3.6B active
  • Phi-4Microsoft · 14.7B

$2,000–5,0004 free · 8 with a pass

Workstation. From the Mac mini M4 Pro 64GB ($2,116) up to the Mac Studio M4 Max 128GB ($3,683) for the largest.

  • Mistral Small 4Mistral · 119B, 6.5B active
  • gpt-oss-120bOpenAI · 117B, 5.1B active
  • Llama 4 ScoutMeta · 109B, 17B active
  • Qwen2.5 72B InstructQwen · 72.7B
  • Llama 3.3 70B InstructMeta · 70.6B
  • Qwen3.6 35B A3BQwen · 35B, 3B active
  • Qwen2.5 Coder 32BQwen · 32B
  • Qwen3 32BQwen · 32B

Over $5,0001 free · 4 with a pass

Pro workstation or company server. From the Mac Studio M5 Ultra 256GB ($13,994) up to the 8× NVIDIA H200 server, about $370,000 (estimate, source) for the largest.

  • Kimi K2.6Moonshot AI · 1000B, 32B active
  • DeepSeek R1 0528DeepSeek · 671B, 37B active
  • DeepSeek V3DeepSeek · 671B, 37B active
  • Qwen3 235B A22BQwen · 235B, 22B active

Online only6 with a pass

GPT, Claude, Gemini.

  • Claude Sonnet 5.5Anthropic
  • GPT-5.6 TerraOpenAI
  • Gemini 3.1 Pro PreviewGoogle
  • Gemini 3.8 FlashGoogle
  • Claude Opus 5.5Anthropic
  • GPT-5.6 SolOpenAI

Every model with a lock needs a pass

Open models in every price class, up to a company server, and GPT, Claude and Gemini. Test them side by side on the full compare.

Get notified when passes open

Passes are not on sale yet. Open the full compare to see the models.