Memory
24 / 17.5 GB
6.5 GB estimated headroom after context allowance.
Can your machine stream the experts?
Shape an honest experiment plan from RAM, NVMe bandwidth, CPU threads, context, and generation goals. This calculator models pressure points; it does not run GLM-5.2 or invent tokens-per-second claims.
Experiment verdict
Ready to experiment
This profile clears the planning thresholds for memory, storage streaming, and CPU scheduling. Start with a short generation target and measure locally.
100fit score
Expert stream
3500 / 2832 MB/s
NVMe target includes context and token-pressure allowance.
CPU schedule
12 / 12 threads
Thread target scales with the requested generation burst.
Primary constraint
No blocking constraint
The profile clears every planning threshold. Benchmark a short prompt first and watch real memory and disk activity.