Inside The Machine
Single-board computer · current · data 2026-10-15

Raspberry Pi 5 (16GB), CPU only for local AI.

With 16 GB it runs K2 Horizon 7B (needs 8.6 GB). 38 models fit. Published speed: llama3.2 3B at 5.55 tokens/sec.

Memory16 GB

system

Bandwidth17 GB/s

Higher means faster answers. Up to ~17 GB/s (32-bit LPDDR4X), per ComputingForGeeks.

PriceFrom about $305

As of 2026-10-12. Check before buying.

Which models fit

By memory option.

Usable: 75% on Macs and shared-memory PCs, VRAM minus 1 GB on graphics cards (estimates).
MemoryUsableModels that fitBest three
16 GB12 GB38K2 Horizon 7B (8.6 GB), Gemma 4 12B (11.2 GB), Qwen3.5 9B (9.6 GB)
Speed

Published speeds.

Measured by others, with sources. Not our tests.

ModelTokens/secEngineSetupSource
llama3.2 3B5.55Ollama · Q4 (Ollama default)Pi 5 8GB CPU (Jeff Geerling)community-measured · 2026-10-12
qwen2 1.5B12.47Ollama · Q4 (Ollama default)Pi 5 8GB CPU (Jeff Geerling)community-measured · 2026-10-12
Questions

Short answers.

What is the best local LLM for the Raspberry Pi 5 (16GB), CPU only?
With 16 GB, K2 Horizon 7B (needs 8.6 GB, AA 20.6). 38 of 124 models fit.
How fast is the Raspberry Pi 5 (16GB), CPU only for local AI?
Published: llama3.2 3B at 5.55 tokens/sec (Ollama, Pi 5 8GB CPU (Jeff Geerling)); qwen2 1.5B at 12.47 tokens/sec (Ollama, Pi 5 8GB CPU (Jeff Geerling)). These are other people's measurements, not ours.
How much does the Raspberry Pi 5 (16GB), CPU only cost?
From about $305 (as of 2026-10-12). 2026 prices are inflated by the memory-chip shortage; check before buying.

Specs from ComputingForGeeks – Jetson Orin Nano Super vs AI HAT+ 2 vs Hailo-10H (prices/benchmarks checked 2026-10-12) (2026-10-12). Prices are approximate and dated; 2026 prices are inflated by the memory-chip shortage.

AGENT VIEW · text/markdown · or send Accept: text/markdown
# Raspberry Pi 5 (16GB), CPU only for local AI

URL: https://insidethemachine.ai/machines/raspberry-pi-5-16gb
Type: Single-board computer. Chip: BCM2712 (4× Cortex-A76). Memory: system, options 16 GB. Bandwidth: 17 GB/s (Up to ~17 GB/s (32-bit LPDDR4X), per ComputingForGeeks.). Spec source: https://computingforgeeks.com/edge-llm-jetson-orin-vs-ai-hat-2-vs-hailo-10h/ (2026-10-12).
Price: from about $305 (2026-10-12, approximate). Status: current.

## Which models fit

| memory GB | usable GB | models that fit | best 3 |
|---|---|---|---|
| 16 | 12 | 38 | K2 Horizon 7B (8.6 GB), Gemma 4 12B (11.2 GB), Qwen3.5 9B (9.6 GB) |

## Published speeds (not our tests)

- llama3.2 3B: 5.55 tok/s decode (Ollama, Q4 (Ollama default), Pi 5 8GB CPU (Jeff Geerling), community-measured) https://computingforgeeks.com/edge-llm-jetson-orin-vs-ai-hat-2-vs-hailo-10h/ (2026-10-12)
- qwen2 1.5B: 12.47 tok/s decode (Ollama, Q4 (Ollama default), Pi 5 8GB CPU (Jeff Geerling), community-measured) https://computingforgeeks.com/edge-llm-jetson-orin-vs-ai-hat-2-vs-hailo-10h/ (2026-10-12)

## Questions

**What is the best local LLM for the Raspberry Pi 5 (16GB), CPU only?**
With 16 GB, K2 Horizon 7B (needs 8.6 GB, AA 20.6). 38 of 124 models fit.

**How fast is the Raspberry Pi 5 (16GB), CPU only for local AI?**
Published: llama3.2 3B at 5.55 tokens/sec (Ollama, Pi 5 8GB CPU (Jeff Geerling)); qwen2 1.5B at 12.47 tokens/sec (Ollama, Pi 5 8GB CPU (Jeff Geerling)). These are other people's measurements, not ours.

**How much does the Raspberry Pi 5 (16GB), CPU only cost?**
From about $305 (as of 2026-10-12). 2026 prices are inflated by the memory-chip shortage; check before buying.