🔌
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 MI430X run Japanese-LM-3.6B?
Yes — here's how fast and up to what context, with real measured numbers.
Japanese-LM-3.6B on the Instinct MI430X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 5932k | 2038.8-2491.8 t/s | ~3.5-4.3 (slow) |
| Q8 | yes | 5987k | 4077.5-4983.6 t/s | ~7-8.6 (slow) |
| Q4 | yes | 6011k | 7281.3-8899.3 t/s | ~12.5-15.3 (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 MI430X run Japanese-LM-3.6B?▶
Yes. At Q4 it generates ~7281.3-8899.3 tok/s and fits up to 6011k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~12.5-15.3 tok/s.
How much context fits?▶
At Q4, up to about 6011k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.