Leaderboard

Real benchmarks from real hardware. Compare local LLM performance across different setups. Run the command below to contribute yours.

$ npm install -g metrillm@latest
$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest

Trending

Last 7 days
Top Models
  • 1
    gemma3:1b 3
  • 2
    deepseek-r1:latest 2
  • 3
    qwen3:4b 2
  • 4
    gemma3:4b 2
  • 5
    gpt-oss:20b 2
Top Machines
  • 1
    Docker Container 24
  • 2
    NVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super 17
  • 3
    NVIDIA NVIDIA_DGX_Spark 14
  • 4
    Mac mini 6
  • 5
    AZW GTR Pro 1
Top CPUs
  • 1
    Intel Gen Intel® Core™ i9-11900H 24
  • 2
    Cortex-A78AE 17
  • 3
    Cortex-X925 14
  • 4
    Apple M4 Pro 6
  • 5
    AMD RYZEN AI MAX+ 395 1

Benchmarks

360

total runs

Models

241

unique models

Families

45

model families

Hardware

31

unique CPUs

Get notified when new models & hardware are benchmarked
Filters

360 results

#CPUModelSizeRAMRuntimetok/sTTFTHW FitQualityGlobalFlagsVerdict
1Apple M2 Maxqwen3-30b-a3b-thinking-2507-claude-4.5-sonnet-high-reasoning-distill-mlx30B/MXFP496 GBLM-STUDIOMLX79.61.7 s
95
98
97
THINKExcellent
2AMD RYZEN AI MAX+ 395nemotron3:33b33.0B/Q4_K_M125 GBOLLAMAGGUF62.71.0 s
95
97
96
ECOTHINKExcellent
3Apple M4 Progpt-oss-20b—64 GBLM-STUDIOGGUF118.22.2 s
94
95
95
THINKExcellent
4Apple M2 Maxqwen3-30b-a3b-thinking-2507-claude-4.5-sonnet-high-reasoning-distill-mlx30B/MXFP496 GBLM-STUDIOMLX80.41.8 s
94
96
95
THINKExcellent
5Intel Core™ Ultra 9 285Kgpt-oss:20b20.9B/MXFP447 GBOLLAMAGGUF239.9616 ms
98
94
95
THINKExcellent
6Apple M4openai/gpt-oss-20b20B/MXFP424 GBLM-STUDIOMLX29.31.8 s
99
93
95
THINKExcellent
7Apple M4 Progpt-oss:20b20.9B/MXFP448 GBOLLAMAGGUF54.52.9 s
92
95
94
THINKExcellent
8Apple M4openai/gpt-oss-20b20B/MXFP432 GBLM-STUDIOMLX40.71.3 s
100
92
94
THINKExcellent
9Apple M4qwen/qwen3-30b-a3b-250730B/4bit32 GBLM-STUDIOMLX44.8405 ms
100
91
94
Excellent
10Cortex-X925gpt-oss:20b20.9B/MXFP4122 GBOLLAMAGGUF60.53.1 s
87
95
93
THINKExcellent
11Apple M4 Pronemotron-3-nano—64 GBLM-STUDIOGGUF93.3257 ms
100
90
93
THINKExcellent
12AMD RYZEN AI MAX+ 395gpt-oss:20b20.9B/MXFP4125 GBOLLAMAGGUF47.11.8 s
85
96
93
ECOTHINKExcellent
13AMD Ryzen 5 5500qwen/qwen3-vl-30b30B/Q4_K_M31 GBLM-STUDIOGGUF24.8735 ms
95
90
92
Excellent
14AMD Ryzen 5 5600X 6-Core Processorqwen3.6-35b-a3b-claude-4.7-opus-reasoning-distilled-apex-mtp35B47 GBLM-STUDIOGGUF42.63.1 s
87
94
92
ECOTHINKExcellent
15Cortex-X925gpt-oss:20b20.9B/MXFP4122 GBOLLAMAGGUF60.73.2 s
86
95
92
THINKExcellent
16Intel Core™ Ultra 9 285HXqwen3-coder:30b-a3b-q4_K_M30.5B/Q4_K_M63 GBOLLAMAGGUF175.3224 ms
100
87
91
Excellent
17Intel Core™ Ultra 9 285Kqwen3-coder:30b30.5B/Q4_K_M47 GBOLLAMAGGUF207.9513 ms
96
87
90
Excellent
18Intel Core™ Ultra 9 285HXqwen3-coder:30b-a3b-q8_030.5B/Q8_063 GBOLLAMAGGUF43.1957 ms
89
90
90
Excellent
19AMD Ryzen 5 5500openai/gpt-oss-20b20B/MXFP431 GBLM-STUDIOGGUF58.74.5 s
88
91
90
THINKExcellent
20Apple M4unsloth/gemma-4-26b-a4b-it26B/Q3_K_M24 GBLM-STUDIOGGUF24.9817 ms
85
92
90
Excellent
...

Don't miss new benchmarks

Get notified when new models and hardware configurations are tested. No spam, unsubscribe anytime.