🔌
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 Instinct MI455X run Qwen2.5-3B?
Yes — here's how fast and up to what context, with real measured numbers.
Qwen2.5-3B on the Instinct MI455X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 5947k | 1998.1-2442.1 t/s | ~4.1-5 (slow) |
| Q8 | yes | 5994k | 3996.1-4884.1 t/s | ~8.2-10 (slow) |
| Q4 | yes | 6015k | 7135.9-8721.7 t/s | ~14.6-17.8 (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 Instinct MI455X run Qwen2.5-3B?▶
Yes. At Q4 it generates ~7135.9-8721.7 tok/s and fits up to 6015k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~14.6-17.8 tok/s.
How much context fits?▶
At Q4, up to about 6015k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.