DefiledAI Tools
GPU PRICE/PERFORMANCE
Current GPU rankings by inference value. Prices updated 2026-05-30. Sorted by tokens per second per dollar by default.
| Rank | GPU | VRAM | Street Price | 7B Tok/s | 70B Tok/s | Tok/s per $100 (7B Q4) |
|---|---|---|---|---|---|---|
| #1 | RTX 3070 Ti Ampere | 8GB | $240 | 62 | N/A | 25.8 |
| #2 | RTX 3070 Ampere | 8GB | $210 | 52 | N/A | 24.8 |
| #3 | RTX 3080 10GB Ampere | 10GB | $320 | 78 | N/A | 24.4 |
| #4 | RTX 3060 12GB Ampere | 12GB | $180 | 42 | N/A | 23.3 |
| #5 | RTX 3080 12GB Ampere | 12GB | $380 | 86 | N/A | 22.6 |
| #6 | RTX 3080 Ti Ampere | 12GB | $420 | 88 | N/A | 21.0 |
| #7 | RX 6800 XT RDNA2 | 16GB | $280 | 52 | N/A | 18.6 |
| #8 | RTX 4070 Ada | 12GB | $370 | 68 | N/A | 18.4 |
| #9 | RX 7800 XT RDNA3 | 16GB | $370 | 62 | N/A | 16.8 |
| #10 | RTX 4070 Super Ada | 12GB | $460 | 74 | N/A | 16.1 |
| #11 | RTX 4070 Ti Ada | 12GB | $530 | 76 | N/A | 14.3 |
| #12 | RTX 3090 Ampere · NVLink | 24GB | $680 | 96 | N/A | 14.1 |
| #13 | RX 7900 XT RDNA3 | 20GB | $540 | 76 | N/A | 14.1 |
| #14 | RTX 4060 Ti 16GB Ada | 16GB | $370 | 48 | N/A | 13.0 |
| #15 | RTX 4070 Ti Super Ada | 16GB | $680 | 88 | N/A | 12.9 |
| #16 | RX 7900 XTX RDNA3 | 24GB | $680 | 88 | N/A | 12.9 |
| #17 | RTX 3090 Ti Ampere · NVLink | 24GB | $820 | 104 | N/A | 12.7 |
| #18 | RTX 4080 Ada | 16GB | $750 | 94 | N/A | 12.5 |
| #19 | RTX 4080 Super Ada | 16GB | $850 | 98 | N/A | 11.5 |
| #20 | RTX 4090 Ada | 24GB | $1,350 | 128 | N/A | 9.5 |
| #21 | 2× RTX 3090 NVLink Multi · NVLink | 48GB | $1,360 | 98 | 21 | 7.2 |
| #22 | 2× RX 7900 XTX Multi | 48GB | $1,360 | 94 | 19 | 6.9 |
| #23 | 2× RTX 4090 Multi | 48GB | $2,700 | 132 | 35 | 4.9 |
Prices are estimated street/used market values as of 2026-05-30. Tok/s measured at Q4_K_M with ExLlamaV2 on NVIDIA, llama.cpp on AMD. Multi-GPU configs assume PCIe unless noted.