We treat '30B-class' as open-weight models with 20–45B parameters — 13 in our catalog. The table ranks every catalog GPU that fully fits at least one of them at Q4 / 8K context, by average estimated decode tok/s across the models it fits. Fit means weights + KV cache + 1 GB runtime overhead within 95% of VRAM — not weights alone. Methodology.
| # | GPU | VRAM (GB) | Bandwidth (GB/s) | MSRP | Fits | Avg Q4 tok/s | Buy |
|---|---|---|---|---|---|---|---|
| 1 | B200 192GB SXM | 192 | 8,000 | $40,000 | 13/13 | 742.3 est. | — |
| 2 | MI300X 192GB | 192 | 5,300 | $15,000 | 13/13 | 491.8 est. | — |
| 3 | H200 141GB SXM5 | 141 | 4,800 | $35,000 | 13/13 | 445.4 est. | — |
| 4 | H100 80GB SXM5 | 80 | 3,350 | $30,000 | 13/13 | 310.8 est. | — |
| 5 | Gaudi 3 128GB | 128 | 3,276 | $12,500 | 13/13 | 304.0 est. | — |
| 6 | RTX 5080 16GB | 16 | 960 | $999 | 1/13 | 204.9 incl. measured | NVIDIA GeForce RTX 5080 Founders Edition (affiliate) · More listings |
| 7 | A100 80GB SXM4 | 80 | 2,039 | $15,000 | 13/13 | 190.5 incl. measured | — |
| 8 | RTX 5070 Ti 16GB | 16 | 896 | $749 | 1/13 | 189.5 incl. measured | Search Amazon (affiliate) |
| 9 | RTX 4080 Super 16GB | 16 | 736 | $999 | 1/13 | 186.5 incl. measured | ASUS ProArt GeForce RTX 4080 Super OC (affiliate) · More listings |
| 10 | RTX 5090 32GB | 32 | 1,792 | $1,999 | 13/13 | 166.7 incl. measured | Search Amazon (affiliate) |
| 11 | RTX 4090 24GB | 24 | 1,008 | $1,599 | 9/13 | 131.6 incl. measured | GIGABYTE GeForce RTX 4090 Gaming OC 24GB (affiliate) · More listings |
| 12 | RTX 3090 Ti 24GB | 24 | 1,008 | $1,999 | 9/13 | 126.4 est. | Search Amazon (affiliate) |
| 13 | RTX 4090D 24GB (China) | 24 | 1,008 | $1,599 | 9/13 | 126.4 est. | — |
| 14 | RX 7900 XTX 24GB | 24 | 960 | $999 | 9/13 | 120.4 est. | ASRock Radeon RX 7900 XTX Phantom Gaming 24GB OC (affiliate) · More listings |
| 15 | RTX 4070 Ti Super 16GB | 16 | 672 | $799 | 1/13 | 118.8 est. | Search Amazon (affiliate) |
| 16 | RTX 3090 24GB | 24 | 936 | $1,499 | 9/13 | 116.9 incl. measured | MSI GeForce RTX 3090 Ventus 3X 24G OC (affiliate) · More listings |
| 17 | RX 9070 16GB | 16 | 640 | $549 | 1/13 | 113.1 est. | Search Amazon (affiliate) |
| 18 | RX 9070 XT 16GB | 16 | 640 | $599 | 1/13 | 113.1 est. | Search Amazon (affiliate) |
| 19 | RTX 5060 Ti 16GB | 16 | 448 | $429 | 1/13 | 111.5 incl. measured | Search Amazon (affiliate) |
| 20 | RX 7900 XT 20GB | 20 | 800 | $899 | 8/13 | 110.5 est. | Search Amazon (affiliate) |
| 21 | RX 7800 XT 16GB | 16 | 624 | $499 | 1/13 | 110.3 est. | Search Amazon (affiliate) |
| 22 | RX 7900 GRE 16GB | 16 | 576 | $549 | 1/13 | 101.8 est. | Search Amazon (affiliate) |
| 23 | Arc A770 16GB | 16 | 560 | $349 | 1/13 | 99.0 est. | Search Amazon (affiliate) |
| 24 | RTX 6000 Ada 48GB | 48 | 960 | $6,800 | 13/13 | 89.1 est. | Search Amazon (affiliate) |
| 25 | L40S 48GB | 48 | 864 | $8,000 | 13/13 | 80.2 est. | Search Amazon (affiliate) |
| 26 | Mac Studio M3 Ultra 96GB | 96 | 819 | $3,999 | 13/13 | 73.7 incl. measured | Search Amazon (affiliate) |
| 27 | Mac Studio M2 Ultra 192GB | 192 | 800 | $6,999 | 13/13 | 72.3 incl. measured | Search Amazon (affiliate) |
| 28 | RTX A6000 48GB | 48 | 768 | $4,500 | 13/13 | 71.3 est. | Search Amazon (affiliate) |
| 29 | T4 16GB | 16 | 320 | $2,000 | 1/13 | 56.6 est. | Search Amazon (affiliate) |
| 30 | RTX 4060 Ti 16GB | 16 | 288 | $499 | 1/13 | 50.9 est. | Search Amazon (affiliate) |
| 31 | RX 7600 XT 16GB | 16 | 288 | $329 | 1/13 | 50.9 est. | Search Amazon (affiliate) |
| 32 | MacBook Pro M4 Max 128GB | 128 | 546 | $4,399 | 13/13 | 50.7 est. | Search Amazon (affiliate) |
| 33 | MacBook Pro M3 Max 128GB | 128 | 400 | $4,199 | 13/13 | 37.1 est. | Search Amazon (affiliate) |
| 34 | Mac mini M4 Pro 64GB | 64 | 273 | $2,199 | 13/13 | 25.3 est. | Search Amazon (affiliate) |
| 35 | Mac mini M4 32GB | 32 | 120 | $999 | 13/13 | 11.1 est. | Search Amazon (affiliate) |