Specs at a glance
| CPU | AMD Ryzen 5 PRO 4650GE / 5650GE |
| RAM options | 8 / 16 / 32 GB DDR4 |
| Storage | NVMe SSD 256 GB to 1 TB |
| Power draw | 15-35 W |
| Form factor | 182 × 179 × 35 mm |
| Local LLM capability | Small models only (≤ 3B Q4) |
| Price point | €150-250 used |
Choosing between devices? The hardware buyer's guide compares them by budget, and Local LLM benchmarks has the published speeds, machine by machine.
Overview
The ThinkCentre M75q (Tiny) is one of the great refurb deals in 2026. €150-250 buys a 2021-era Ryzen 5 with 16 GB RAM, an NVMe slot, USB 3.0+: comfortably running Hermes Agent with browser tool. Local models are the limit: on llama.cpp's Vulkan scoreboard, the Ryzen 4000 series (the 4650GE's family; the scoreboard names the family, not the chip) writes 9.63 tokens per second with a 7B model at Q4_0. For an agent that calls a cloud model, the price is hard to beat. Watch the seller for genuine refurb units versus “tested working” ones with worn fans.
Best for
- Budget self-hosters wanting more than a Pi 5 can offer
- Cost-optimised always-on agent hosting
- Refurb-friendly builders
Not for
- Local LLM beyond 7B Q4 (modest GPU)
- Brand-new-hardware purists
- Specific compliance requirements (used hardware introduces audit gaps)
Compatible self-hosted agents
Listed as compatible with Lenovo ThinkCentre M75q (used) from each agent's requirements, not from a test (with the caveats from “Best for” / “Not for” above):
Where to buy
We don't have an active affiliate programme with this vendor: see our disclosure page for the full list of partners we do work with.
See also: all AI hardware · AI agents tracker · AI CVEs · edge AI hardware buyer's guide · scan your AI stack · our methodology.