Models
Which model do you want to run?
Pick a model and see the bare minimum to run it, every GPU configuration that fits, AMD against NVIDIA, and what a fleet of a thousand costs, all from the model's own config and live rental prices.
56 models, 105 datacenter GPUs, refreshed with every release.
Popular models
All models
56models
| 2.78 T | 16+ | Explore → | |
| 2.45 T | 16+ | Explore → | |
| 1.60 T | 16 | Explore → | |
| 1.03 T | 8 | Explore → | |
| 1.03 T | 8 | Explore → | |
| 1.03 T | 16 | Explore → | |
| 1.02 T | 16 | Explore → | |
| 753 B | 16 | Explore → | |
| 753 B | 16+ | Explore → | |
| 685 B | 16 | Explore → | |
| 684 B | 16 | Explore → | |
| 561 B | 16 | Explore → | |
| 427 B | 16 | Explore → | |
| 406 B | 16 | Explore → | |
| 403 B | 16 | Explore → | |
| 402 B | 16 | Explore → | |
| 358 B | 16 | Explore → | |
| 358 B | 16 | Explore → | |
| 311 B | 8 | Explore → | |
| 291 B | 4 | Explore → | |
| 253 B | 8 | Explore → | |
| 235 B | 8 | Explore → | |
| 229 B | 4 | Explore → | |
| 229 B | 4 | Explore → | |
| 229 B | 4 | Explore → | |
| 199 B | 8 | Explore → | |
| 180 B | 8 | Explore → | |
| 128 B | 2 | Explore → | |
| 125 B | 4 | Explore → | |
| 124 B | 4 | Explore → | |
| 117 B | 1 | Explore → | |
| 109 B | 4 | Explore → | |
| 106 B | 4 | Explore → | |
| 79.7 B | 4 | Explore → | |
| 70.6 B | 2 | Explore → | |
| 70.6 B | 2 | Explore → | |
| 49.9 B | 2 | Explore → | |
| 36.0 B | 1 | Explore → | |
| 32.9 B | 1 | Explore → | |
| 32.8 B | 1 | Explore → | |
| 32.2 B | 1 | Explore → | |
| 32.2 B | 1 | Explore → | |
| 31.6 B | 1 | Explore → | |
| 31.3 B | 1 | Explore → | |
| 31.2 B | 1 | Explore → | |
| 27.8 B | 1 | Explore → | |
| 27.8 B | 1 | Explore → | |
| 27.8 B | 1 | Explore → | |
| 27.4 B | 1 | Explore → | |
Gemma 4 26B-A4B MoE | 25.8 B | 1 | Explore → |
| 24.0 B | 1 | Explore → | |
| 20.9 B | 1 | Explore → | |
| 14.7 B | 1 | Explore → | |
| 8.9 B | 1 | Explore → | |
| 8.0 B | 1 | Explore → | |
| 4.0 B | 1 | Explore → |
Weights at each model's native precision. GPU counts are the tensor-parallel width of
H100 SXM5 80GB parts: minimum is 4,096 tokens of context and one stream, recommended is 32,768
tokens and 32 streams. Computed from each model's config.json.