Models
Speed is not a model property. Rank is the fastest measured pair of that family.
| # | Model · Build | Hardware | Decode · tok/s | Get |
|---|---|---|---|---|
RTX 5090 | 370 | |||
RTX 5090 32GB | 351 | |||
4× RTX PRO 6000 | 264 | |||
RTX 4090 24GB | 220 | |||
RTX 5090 | 195.8 | |||
RTX 5090 32GB | 173 | |||
RTX 5060 Ti 16GB + RTX 3070 8GB | 127 | |||
M4 Pro 64GB | 108.6 | |||
2× DGX Spark | 103 | |||
2× DGX Spark | 94.4 | |||
Radeon RX 9060 XT 16GB | 88 | |||
M4 Pro | 82.8 | |||
2× DGX Station GB300 | 81.5 | |||
DGX Spark | 80 | |||
M5 Max 128GB | 65 | |||
DGX Spark | 54.7 | |||
RTX 5060 Ti 16GB | 54.7 | |||
DGX Spark | 53 | |||
RTX 3060 12GB | 50 | |||
M4 Pro | 49.3 | |||
M4 Pro | 48.3 | |||
RTX 5060 8GB | 47 | |||
GTX 1660 SUPER 6GB | 40 | |||
DGX Spark | 33.0 | |||
RTX 5060 Ti 16GB | 14 | |||
RTX 5060 Ti 16GB | 10 | |||
Radeon RX 9070 XT | 10 | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
| — | — | |||
Qwen-Image-2.1 UncensoredUncensored | — | — | ||
| — | — |








