LFM2.5 8B-A1B
LFM2.5 8B-A1B loads in 5.5 GB at Q4_K_M and decodes at the speed of its 1.5B active path, comfortably from 12 GB of memory — Macs included, starting at the Mac Mini M6 16GB. Below is the memory math we ran for it, the cheapest card that fits, and the fastest machine we track.
How much memory does LFM2.5 8B-A1B need?
LFM2.5 8B-A1B is a 8.3B mixture-of-experts model (1.5B active per token) built for on-device agents, tool calling, multilingual chat. The weights alone take 5.5 GB at Q4_K_M; add context and the OS, and it belongs in the 12 GB memory class. Every number below comes from the ModelFit engine — speeds are estimates, labeled est., because we do not benchmark hardware ourselves.
KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).
Which machines run LFM2.5 8B-A1B?
Verdicts come from the same engine that powers our fit checker — memory capacity first, bandwidth for speed. How we compute these numbers · Hugging Face model card
More LFM2 models
Frequently asked questions
How much memory does LFM2.5 8B-A1B need?
5.5 GB for the Q4_K_M weights, plus ~2.0 GB of KV-cache at 16k context — about 7.5 GB total. Comfortable from 12 GB of VRAM or unified memory.
Does LFM2.5 8B-A1B run on a Mac?
Yes — from the Mac Mini M6 16GB (~64 tok/s est.). Unified memory means the RAM budget is the only limit.
What is the cheapest GPU for LFM2.5 8B-A1B?
The AMD Radeon RX 7900 XT is the cheapest tracked card that runs LFM2.5 8B-A1B comfortably — ~148 tok/s est. at ~$550 used.
Cite this page
ModelFit: LFM2.5 8B-A1B — specs, memory math and hardware verdicts. https://modelfit.io/models/lfm2.5-8b-a1b/ (dataset updated 2026-09-03, CC BY 4.0).