Best graphics cards for local AI under $2,000
Ranked by what each one runs fully in memory at 32k context, with prices in the US today.
In short
The top pick in the US right now is the NVIDIA GeForce RTX 3090, $1,600 at eBay: it runs Qwen 3.8 27B fully in memory at 32k context at about 26 tok/s (estimated). Ranked by the most capable open model each one runs fully in memory at 32k context and a usable speed, then by how many models it runs, then by speed on Qwen 3.8 27B, then by price.
Updated . Speeds are estimates; how they are worked out.
The list
Best first. Speeds are estimates est.
| # | Card | Memory | Price | Best model it runs at 32k | Models in memory | Qwen 3.8 27B | tok/s per $1,000 |
|---|---|---|---|---|---|---|---|
| 1 | NVIDIA GeForce RTX 3090 | 24 GB | $1,600eBay | Qwen 3.8 27BUD-Q5_K_M26 tok/s | 26 | 26 tok/s | 16.4 |
| 2 | AMD Radeon RX 7900 XTX | 24 GB | $1,533Amazon | Qwen 3.8 27BUD-Q5_K_M24 tok/s | 26 | 24 tok/s | 15.6 |
| 3 | AMD Radeon RX 7900 XT | 20 GB | $800Amazon | Qwen 3.8 27BUD-Q4_K_M24 tok/s | 18 | 24 tok/s | 29.4 |
| 4 | AMD Radeon AI PRO R9700 | 32 GB | $1,700Amazon | Qwen 3.6 35B-A3BUD-Q6_K66 tok/s | 20 | 11 tok/s | 6.6 |
| 5 | Intel Arc Pro B60 | 24 GB | $650Amazon | Qwen 3.6 35B-A3BUD-Q4_K_M49 tok/s | 16 | 10 tok/s | 15.8 |
| 6 | NVIDIA Tesla P40 | 24 GB | $405eBay | Qwen 3.6 35B-A3BUD-Q4_K_M50 tok/s | 15 | 10 tok/s | 24 |
| 7 | Intel Arc A770 16GB | 16 GB | $400Amazon | gpt-oss 20BMXFP448 tok/s | 13 | offload | – |
| 8 | AMD Radeon RX 7800 XT | 16 GB | $550eBay | gpt-oss 20BMXFP464 tok/s | 13 | offload | – |
| 9 | NVIDIA GeForce RTX 5060 Ti 16GB | 16 GB | $618eBay | gpt-oss 20BMXFP452 tok/s | 13 | offload | – |
| 10 | AMD Radeon RX 9070 XT | 16 GB | $815Amazon | gpt-oss 20BMXFP465 tok/s | 13 | offload | – |
| 11 | NVIDIA GeForce RTX 5070 Ti | 16 GB | $995eBay | gpt-oss 20BMXFP4103 tok/s | 13 | offload | – |
| 12 | NVIDIA GeForce RTX 4070 Ti Super | 16 GB | $1,013Amazon | gpt-oss 20BMXFP477 tok/s | 13 | offload | – |
Questions
- What is the best graphics card for running AI models locally under $2,000?
- The top pick in the US right now is the NVIDIA GeForce RTX 3090, $1,600 at eBay: it runs Qwen 3.8 27B fully in memory at 32k context at about 26 tok/s (estimated).
- Which gives the most speed for the money?
- The AMD Radeon RX 7900 XT: about 24 tok/s on Qwen 3.8 27B for $800, the most tokens per second per unit of price on this list (estimated).
- What is the cheapest option here that runs Qwen 3.8 27B fully in memory?
- The NVIDIA Tesla P40, $405 at eBay, at about 10 tok/s at 32k context (estimated).