What runs this LLM?
Pick the model you want to serve. We size it from the released checkpoint and list the cheapest configurations you can rent today that hold it — with the on-demand and serverless rate for each.
1 · Choose a model
2 · What it needs
DeepSeek R1 at 8-bit
826 GB of VRAM
at short context — measured from the released checkpoint, not estimated from parameter count.
- Model weights
- 689 GB
- KV cachefolded into overhead
- —
- Runtime overhead20% of weights
- 138 GB
About this model
- Parameters
- 671B
- Active per token
- 37B
- Released at
- 8-bit
- Measured weights
- 689 GB
3 · Where to run it
Bookable configurations
Cheapest first, checked against 98 GPU types with live bookable capacity.
| Configuration | Total VRAM | On-demand | Serverless | GPU details |
|---|---|---|---|---|
| 8× NVIDIA RTX PRO 6000Tight fit | 768 GB 58 GB short of the margin | $8.00/hr $1.00/GPU-hr  Lium | $16.63/hr $2.08/GPU-hr  Verda | See all prices |
| 8× Intel Gaudi 2Tight fit | 768 GB 58 GB short of the margin | $8.31/hr $1.04/GPU-hr  Cyfuture AI | — | See all prices |
| 4× AMD Instinct MI325X | 1,024 GB 198 GB free | $12.35/hr $3.09/GPU-hr Cyfuture AI | — | See all prices |
| 8× AMD Instinct MI300X | 1,536 GB 710 GB free | $22.48/hr $2.81/GPU-hr Cyfuture AI | — | See all prices |
| 4× NVIDIA B300 | 1,152 GB 326 GB free | $30.00/hr $7.50/GPU-hr Verda | $33.00/hr $8.25/GPU-hr Verda | See all prices |
| 4× NVIDIA B300 CC | 1,152 GB 326 GB free | $30.60/hr $7.65/GPU-hr Verda | — | See all prices |
| 8× NVIDIA H200 | 1,128 GB 302 GB free | $32.00/hr $4.00/GPU-hr Verda | $35.20/hr $4.40/GPU-hr Verda | See all prices |
| 4× NVIDIA GB300 | 1,152 GB 326 GB free | $34.48/hr $8.62/GPU-hr Verda | — | See all prices |
Showing the 8 cheapest. Model weights plus 20% for runtime overhead, from measured checkpoint sizes; a stated context length adds its KV cache on top. Each price is the cheapest bookable rate for that exact instance size, for the whole configuration — the cheaper of the two is highlighted and orders the table. A ×N figure is total VRAM across N cards and makes no claim about interconnect throughput.
How this is worked out
Weight sizes are measured from the actual checkpoint files on Hugging Face — never derived from parameter count, which gets quantized releases wrong. On top of the weights we assume 20% for activations, CUDA context and allocator slack. Stating a context length adds its KV cache explicitly, computed only for architectures whose cache shape the config states unambiguously.
A configuration that holds the weights and cache but not that full margin is marked as a tight fit rather than hidden — it will load, and may well run at batch size one, but expect trouble under load.
Prices are the cheapest bookable rate for that exact instance size, refreshed with the rest of the catalog.