🔌
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 H200 run ImageBind?
Yes — here's how fast and up to what context, with real measured numbers.
ImageBind on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1928k | 1879.8-2240.3 t/s | ~15.7-18.7 (slow) |
| Q8 | yes | 1942k | 3759.7-4480.7 t/s | ~31.3-37.3 (slow) |
| Q4 | yes | 1949k | 6713.7-8001.2 t/s | ~55.9-66.7 (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.
FAQ
Can the NVIDIA H200 run ImageBind?▶
Yes. At Q4 it generates ~6713.7-8001.2 tok/s and fits up to 1949k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~55.9-66.7 tok/s.
How much context fits?▶
At Q4, up to about 1949k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.