Raspberry Pi 5 + AI HAT+ 2 / Llama 3.2 3B

Can a Raspberry Pi 5 + AI HAT+ 2 run Llama 3.2 3B?

Technically yes, but slowly. Measured generation speed: 1.01–2.65 tok/s across 3 public runs.

Rule of thumb: 10+ tok/s feels like a chat; 4–10 is usable for assistants and automations; under 4 is batch-job territory. Speeds depend on quantization, runtime, context length and cooling. Check each source.

BoardQuantRuntimeGen tok/sPrompt tok/sTTFT msSource
Raspberry Pi 5 + AI HAT+ 2
Raspberry Pi 5 8GB + AI HAT+ 2
Hailo-compiledhailo-ollama2.65––source ↗
tested 2026-01-25
Raspberry Pi 5 + AI HAT+ 2
Raspberry Pi 5 + AI HAT+ 2
Hailo-compiledhailo-ollama2.6––source ↗
2026-01-20
Raspberry Pi 5 + AI HAT+ 2
Raspberry Pi 5 + AI HAT+ 2
not reportedhailo-ollama1.01––source ↗
2026-04-24

Llama 3.2 3B on other boards

Buy a Raspberry Pi 5 + AI HAT+ 2

Ran Llama 3.2 3B on a Raspberry Pi 5 + AI HAT+ 2? Submit your numbers →

Get the monthly board report

New boards, new runtimes, new numbers. One email a month, no hype.