Sizing: which model on which machine
This section puts 7 open models against the 8 platforms in the catalogue.
Which model fits on which machine?
Each cell gives the most precise format that loads the model on that machine, and links to the detail page with the footprint in all four formats and the remaining margin.
| Model | DGX Spark | Mac Studio Ultra | DGX Station | RTX PRO 6000 server | H200 SXM server | B200 SXM | B300 SXM | GB300 NVL72 |
|---|---|---|---|---|---|---|---|---|
| GLM 5.2 | does not fit | NVFP4 | NVFP4 | NVFP4 | FP8 | FP8 | FP16 | FP16 |
| Kimi K3 | does not fit | does not fit | does not fit | does not fit | does not fit | does not fit | NVFP4 | FP16 |
| Kimi K2.7 Code | does not fit | does not fit | NVFP4 | NVFP4 | NVFP4 | FP8 | FP16 | FP16 |
| DeepSeek V4 | does not fit | does not fit | does not fit | does not fit | NVFP4 | NVFP4 | FP8 | FP16 |
| Nemotron 3 Ultra | does not fit | NVFP4 | FP8 | FP8 | FP8 | FP16 | FP16 | FP16 |
| MiniMax M3 | does not fit | FP8 | FP8 | FP8 | FP16 | FP16 | FP16 | FP16 |
| Qwen 3.8 27B | FP16 | FP16 | FP16 | FP16 | FP16 | FP16 | FP16 | FP16 |
How to read these figures
The footprint covers the model weights. In a mixture-of-experts architecture every expert stays resident in memory even though only a fraction computes on each token: active parameters govern speed, not occupancy. The attention cache sits on top and grows with context and with the number of concurrent requests.
Method and sources
Footprints are calculated: billion parameters × bytes per parameter of the format × 1.15 runtime margin. They are not measured on hardware. The attention cache sits on top and depends on context and concurrency. Prices are indicative and not contractual. Each detail page carries its sources; the shared conventions are these.
- Bytes per parameter: FP16 2 (16 bits per parameter, hence 2 bytes); FP8 1 (8 bits per parameter (E4M3), hence 1 byte, block scales not counted); NVFP4 0.5 (4 bits per parameter, hence 0.5 byte; 4-bit published weights weigh 0.54 to 0.56 byte per parameter with their scales (DeepSeek V4 Pro 865 GB for 1,599 billion, Kimi K3 1,561 GB for 2,780 billion)); Q4 0.6 (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). Source: connaissance/faits.yaml, quantifications family, checked on 2026-08-31.
- Runtime margin × 1.15: QDNA operating assumption, not measured: 15% above the weights for activations, buffers and fragmentation.
- “Fits” threshold at 75% of memory: QDNA assumption, not measured, which keeps the remaining quarter for the attention cache and concurrency.