What hardware do you need to run LFM2.5 2.6B?

LFM2.5 2.6B at Q4_K_M is 1.7 GB of weights, with a context ceiling of 128k tokens. Not yet rated: released after our last ratings pass. Convolution hybrid, so the context costs almost no memory. A useful floor for small machines.

Cheapest machine that runs itMac mini M6, 16GB at $899

Fastest of the ones listedMac Studio M5 Ultra, 96GB — 408 tok/s at 32k context

Honest answer on costAgainst the API, buying hardware for this model never pays for itself at ordinary usage.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 8, which puts it in the Below every hosted tier band. Fine for simple, well-specified tasks. Noticeably less capable than anything the big labs sell today. Score source. See the whole table.

Summarising — not rated
Translation — not rated
Everyday coding — not rated
Reasoning & maths — not rated
Agentic work — not rated

What it costs either way

Renting the same model costs $0 per million input tokens and $0 per million output (OpenRouter, cheapest active endpoint, checked 2026-09-09). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill.

Machines that run it

Machine	Price	Speed at 32k	Pay-back
Mac mini M6, 16GB	$899	52 tok/s estimated	Never pays back	Run the numbers
Strix Halo Framework Desktop, 32GB	$1,269	87 tok/s estimated	Never pays back	Run the numbers
MacBook Air M5 (13-inch), 16GB	$1,299	52 tok/s estimated	Never pays back	Run the numbers
MacBook Pro M5 (14-inch), 16GB	$1,999	52 tok/s estimated	Never pays back	Run the numbers
Mac Studio M5 Max, 36GB	$2,499	156 tok/s estimated	Never pays back	Run the numbers
DGX Spark GB10 Grace Blackwell, 128GB	$4,699	93 tok/s estimated	Never pays back	Run the numbers

One machine per family, cheapest first. Speeds are measured where a public benchmark exists and estimated from memory bandwidth otherwise; the calculator says which for any configuration.

The specifics

Parameters: 2.7B
Quantisation: Q4_K_M
Weights on disk: 1.7 GB
KV cache: 0.5 GB at 32k context — Only 8 of 30 layers are attention; the rest are short convolutions that keep a fixed-size state. head_dim derived from hidden_size ÷ heads, as the config omits it.
Maximum context: 128k tokens (128k)
Licence: LFM 1.0 (see repo)
Sources: source 1, source 2