🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the NVIDIA RTX PRO 6000 Blackwell run Cosmos-Predict2-14B-Text2Image?
Yes — here's how fast and up to what context, with real measured numbers.
Cosmos-Predict2-14B-Text2Image on the NVIDIA RTX PRO 6000 Blackwell
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 359k | 46.7-55.7 t/s | ~1-1.2 (slow) |
| Q8 | yes | 444k | 93.4-111.4 t/s | ~2.1-2.5 (slow) |
| Q4 | yes | 482k | 166.9-198.9 t/s | ~3.7-4.4 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.
→ All models on the NVIDIA RTX PRO 6000 Blackwell
→ Cosmos-Predict2-14B-Text2Image specs
→ Other hardware
FAQ
Can the NVIDIA RTX PRO 6000 Blackwell run Cosmos-Predict2-14B-Text2Image?▶
Yes. At Q4 it generates ~166.9-198.9 tok/s and fits up to 482k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~3.7-4.4 tok/s.
How much context fits?▶
At Q4, up to about 482k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.