← CanIRunLocal · All hardware · All models
Yes. LFM2.5 8B-A1B (MoE) fits in fast memory on M4 Max (32c GPU) 36GB at Q8_0 (9.8 GB needed, 27.0 GB usable), about 154 tok/s (estimate).
Hardware: M4 Max (32c GPU) 36GB · Model: LFM2.5 8B-A1B (MoE)
M4 Max (32c GPU) 36GB has 36 GB of unified memory and 410 GB/s of memory bandwidth. Mac Studio M4 Max (32c GPU) 36GB has the same memory and bandwidth, so the fit numbers match; this page is M4 Max (32c GPU) 36GB. The checker leaves macOS a share of unified memory, so Ollama can use 27.0 GB. LFM2.5 8B-A1B (MoE) (8.47B, 1B active, Liquid AI) at Q8_0 is a 9 GB download and needs 9.8 GB with headroom for about 8k tokens.
| Best quant | Q8_0 |
|---|---|
| Download size | 9 GB |
| Memory needed | 9.8 GB |
| Usable fast memory | 27.0 GB |
| Fit | fits in fast memory |
| Estimated speed | ~154 tok/s (estimate) |
| Command | ollama run lfm2.5:8b-a1b-q8_0 |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-23.
A project by OnDevice Labs · llms.txt