Edge LLM Leaderboard: Raspberry Pi 5 - Decode — leaderboard

Metric: Decode (tokens/s). Source: huggingface.co. 128 models tracked.

Top models

#ModelScore
1gpt2 (Q4_0_4_4, llama_cpp)101.51
2gpt2 (Q4_K_M, llama_cpp)92.71
3SmolLM2-135M-Instruct (Q4_0_4_4, llama_cpp)86.74
4SmolLM2-135M-Instruct (Q4_K_M, llama_cpp)73.27
5gpt2 (Q8_0, llama_cpp)68.19
6SmolLM2-135M-Instruct (Q8_0, llama_cpp)57.99
7gpt2-medium (Q4_0_4_4, llama_cpp)37.89
8SmolLM2-360M-Instruct (Q4_0_4_4, llama_cpp)37.18
9gpt2-medium (Q4_K_M, llama_cpp)33.91
10SmolLM2-360M-Instruct (Q4_K_M, llama_cpp)30.72

Interactive version: theaggregate.ai/benchmark?slug=edge-llm-leaderboard-raspberry-pi-5-decode · How the rankings work · Data refreshed daily, snapshot 2026-07-22.