bestmodel.run / model — gemma-4-12B-it (agent view) category chat · google/gemma-4-12B-it · 38 runs in the model record answer: 45.2 tok/s (model median) Honesty ladder: measured > reported > extrapolated > formula > no data yet. Empty fields stay empty — never filled with an estimate. rig | quant | median result | basis | ttft | peak vram | max context M1 Max 64GB | 16-bit | 14 tok/s | reported | ttft 2,324ms | vram - GB | ctx 794 Radeon 890M 32GB | 4-bit | 10.4 tok/s | reported | ttft -ms | vram - GB | ctx 4,096 Radeon 890M 32GB | 6-bit | 8.5 tok/s | reported | ttft -ms | vram - GB | ctx 4,096 Radeon 890M 32GB | 8-bit | 6.8 tok/s | reported | ttft -ms | vram - GB | ctx 4,096 Radeon AI Pro R9700 32GB ×3 | 8-bit | 36.7 tok/s | reported | ttft 454ms | vram 15.3 GB | ctx 794 Radeon AI Pro R9700 32GB ×3 | 16-bit | 19.3 tok/s | reported | ttft 1,448ms | vram - GB | ctx 794 RTX 3060 12GB | 4-bit | 76.7 tok/s | measured | ttft 589ms | vram 9.8 GB | ctx 262,144 RTX 3090 24GB | 4-bit | 52.3 tok/s | measured | ttft 4,752ms | vram 0 GB | ctx 131,072 RTX 5000 16GB | 6-bit | 21.5 tok/s | reported | ttft -ms | vram 14 GB | ctx 262,144 RTX 5000 16GB | 8-bit | 23.4 tok/s | reported | ttft -ms | vram 14.9 GB | ctx 98,304 RTX 5090 32GB | 8-bit | 87.2 tok/s | reported | ttft 85ms | vram 13.7 GB | ctx 8,192 RTX PRO 6000 Blackwell 96GB | 16-bit | 86.9 tok/s | reported | ttft 44ms | vram - GB | ctx 2,048 RTX PRO 6000 Blackwell 96GB ×2 | 16-bit | 132.2 tok/s | reported | ttft 29ms | vram - GB | ctx 2,048 RTX PRO 6000 Blackwell 96GB ×4 | 16-bit | 145.7 tok/s | reported | ttft 35ms | vram - GB | ctx 2,048 RX 570 4GB ×2 | 4-bit | 18.9 tok/s | measured | ttft 642ms | vram - GB | ctx 2,048 RX 6700 XT 12GB | 4-bit | 37.8 tok/s | measured | ttft -ms | vram - GB | ctx 4,736