Can a M5Stack Module LLM run DeepSeek-R1-Distill-Qwen 1.5B?
Technically yes, but slowly. Measured generation speed: 2.42–3.57 tok/s across 2 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.
| Board | Size | Runtime | Gen tok/s | Source |
|---|---|---|---|---|
| M5Stack Module LLM M5Stack Module LLM / LLM630 (AX630C) | 1.5B | StackFlow | 3.57 | source ↗ accessed 2026-10-17 |
| M5Stack Module LLM M5Stack Module LLM (AX630C) | 1.5B | StackFlow | 2.42 | source ↗ 2026-04-24 |
DeepSeek-R1-Distill-Qwen 1.5B on other boards
- M5Stack LLM8850 13.29 tok/s best
- Rockchip RK3588 11.5 tok/s best
- Jetson Orin Nano Super 9.59 tok/s best
- Raspberry Pi 5 9.04 tok/s best
- Raspberry Pi 5 + AI HAT+ 2 7.05 tok/s best
Buy a M5Stack Module LLM
- M5Stack LLM Module Kit (AX630C)$79.9regular price; out of stock when checked · price source ↗ (2026-10-17)Check price on Amazon
- M5Stack Module LLM (as tested in paper)$99.9price quoted in arXiv paper · price source ↗ (2026-04-24)Check price on Amazon
Ran DeepSeek-R1-Distill-Qwen 1.5B on a M5Stack Module LLM? Submit your numbers →