# Tesla P40 24GB (used) for local AI

URL: https://insidethemachine.ai/machines/tesla-p40
Type: Used server card. Chip: Tesla P40 24GB (used). Memory: vram, options 24 GB. Bandwidth: 345.6 GB/s. Spec source: https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units (2026-10-15).
Price: from about $277 (2026-10-02, approximate). Status: used.

## Which models fit

| memory GB | usable GB | models that fit | best 3 |
|---|---|---|---|
| 24 | 23 | 56 | Qwen3.8 27B (22.7 GB), Qwen3.5 27B (21.5 GB), Qwen3.6 27B (21.5 GB) |

## Published speeds

None found yet.

Notes: Passive cooling (needs ducted fan), no display out, weak FP16; llama.cpp still supports it. eBay median of 26 sales, range $200–$359. TDP not captured from a fetched source (gap).

## Questions

**What is the best local LLM for the Tesla P40 24GB (used)?**
With 24 GB, Qwen3.8 27B (needs 22.7 GB, AA 33.7). 56 of 124 models fit.

**How much does the Tesla P40 24GB (used) cost?**
From about $277 (as of 2026-10-02). 2026 prices are inflated by the memory-chip shortage; check before buying.