Leaderboard
Real benchmarks from real hardware. Compare local LLM performance across different setups. Run the command below to contribute yours.
npx metrillm@latest benchRequires Node 20+ and Ollama running
Trending
Last 7 days- 1 gemma3:1b 4
- 2 llama3.1:8b 3
- 3 smollm2:360m 2
- 4 phi4:14b 2
- 5 gemma3:12b 2
- 1 ollama 88
- 1 Apple M4 85
- 2 Apple M4 Pro 3
Benchmarks
88
total runs
Models
58
unique models
Families
14
model families
Hardware
2
unique CPUs
Filters
88 results
| # | CPU | Model | Size | RAM | Runtime | tok/s | TTFT | HW Fit | Quality | Global | Flags | Verdict | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Apple M4 | qwen2.5:7b | 7.6B/Q4_K_M | 32 GB | OLLAMAGGUF | 22.2 | 285 ms | 95 | 79 | 85 | Excellent | |||
| 2 | Apple M4 | gemma3:4b | 4.3B/Q4_K_M | 32 GB | OLLAMAGGUF | 34.6 | 303 ms | 100 | 71 | 83 | Excellent | |||
| 3 | Apple M4 | qwen2.5:14b | 14.8B/Q4_K_M | 32 GB | OLLAMAGGUF | 11.2 | 539 ms | 80 | 84 | 82 | Excellent | |||
| 4 | Apple M4 | qwen2.5-coder:7b | 7.6B/Q4_K_M | 32 GB | OLLAMAGGUF | 22.1 | 285 ms | 95 | 72 | 81 | Excellent | |||
| 5 | Apple M4 | yi:6b | 6B/Q4_0 | 32 GB | OLLAMAGGUF | 27.6 | 210 ms | 100 | 67 | 80 | Excellent | |||
| 6 | Apple M4 | qwen2.5:3b | 3.1B/Q4_K_M | 32 GB | OLLAMAGGUF | 47.2 | 168 ms | 100 | 67 | 80 | Excellent | |||
| 7 | Apple M4 | gemma3:12b | 12.2B/Q4_K_M | 32 GB | OLLAMAGGUF | 12.7 | 560 ms | 83 | 78 | 80 | Excellent | |||
| 8 | Apple M4 | gemma2:2b | 2.6B/Q4_0 | 32 GB | OLLAMAGGUF | 55.1 | 152 ms | 100 | 67 | 80 | Excellent | |||
| 9 | Apple M4 | gemma2:9b | 9.2B/Q4_0 | 32 GB | OLLAMAGGUF | 17.5 | 347 ms | 89 | 71 | 78 | Good | |||
| 10 | Apple M4 | llama3.2:3b | 3.2B/Q4_K_M | 32 GB | OLLAMAGGUF | 44.1 | 178 ms | 100 | 63 | 78 | Good | |||
| 11 | Apple M4 Pro | llama3.2:latest | 3.2B/Q4_K_M | 64 GB | OLLAMAGGUF | 98.9 | 125 ms | 100 | 61 | 77 | Good | |||
| 12 | Apple M4 | llama3.1:8b | 8.0B/Q4_K_M | 32 GB | OLLAMAGGUF | 20.3 | 315 ms | 93 | 67 | 77 | Good | |||
| 13 | Apple M4 | granite3.1-dense:8b | 8.2B/Q4_K_M | 32 GB | OLLAMAGGUF | 18.8 | 308 ms | 91 | 66 | 76 | Good | |||
| 14 | Apple M4 Pro | mistral:latest | 7.2B/Q4_0 | 64 GB | OLLAMAGGUF | 54.3 | 124 ms | 100 | 60 | 76 | Good | |||
| 15 | Apple M4 | mistral:latest | 7.2B/Q4_K_M | 32 GB | OLLAMAGGUF | 22.5 | 257 ms | 96 | 63 | 76 | Good | |||
| 16 | Apple M4 | qwen3:0.6b | 751.63M/Q4_K_M | 32 GB | OLLAMAGGUF | 143.9 | 1.4 s | 100 | 59 | 75 | THINK | Good | ||
| 17 | Apple M4 | phi4:14b | 14.7B/Q4_K_M | 32 GB | OLLAMAGGUF | 11.0 | 532 ms | 79 | 73 | 75 | Good | |||
| 18 | Apple M4 | mistral:7b | 7.2B/Q4_K_M | 32 GB | OLLAMAGGUF | 22.0 | 261 ms | 95 | 62 | 75 | Good | |||
| 19 | Apple M4 | phi4-mini:latest | 3.8B/Q4_K_M | 32 GB | OLLAMAGGUF | 36.2 | 190 ms | 100 | 58 | 75 | Good | |||
| 20 | Apple M4 | codegemma:7b | 9B/Q4_0 | 32 GB | OLLAMAGGUF | 18.7 | 320 ms | 91 | 62 | 74 | Good |
Don't miss new benchmarks
Get notified when new models and hardware configurations are tested. No spam, unsubscribe anytime.