🔌
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 Ling-lite-1.5 ("Bailing")?
Yes — here's how fast and up to what context, with real measured numbers.
Ling-lite-1.5 ("Bailing") on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 361k | 104.3-124.3 t/s | ~0.9-1 (slow) |
| Q8 | yes | 425k | 208.6-248.6 t/s | ~1.7-2.1 (slow) |
| Q4 | yes | 453k | 372.4-443.9 t/s | ~3.1-3.7 (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 Ling-lite-1.5 ("Bailing")?▶
Yes. At Q4 it generates ~372.4-443.9 tok/s and fits up to 453k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~3.1-3.7 tok/s.
How much context fits?▶
At Q4, up to about 453k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.