What hardware do you need to run Muse Glimmer 30B?

Muse Glimmer 30B at Q4_K_M is 17 GB of weights, with a context ceiling of 128k tokens. Not yet rated: released after our last ratings pass. Meta's dense 30B under Apache 2.0, aimed at local agents.

Cheapest machine that runs itStrix Halo Framework Desktop, 32GB at $1,269

Shortest pay-backMac mini M6, 32GB — Pays back in 22 years

Fastest of the ones listedMac Studio M5 Ultra, 96GB — 50 tok/s at 32k context

Honest answer on costPays back in 22 years at 500k tokens a day.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 18 (high effort), which puts it in the Haiku-class band. In the same band as Anthropic's cheap, fast tier. Every current OpenAI model scores above this band. 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 $0.3 per million input tokens and $1.1 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
Strix Halo Framework Desktop, 32GB	$1,269	11 tok/s estimated	Pays back in 22 years	Run the numbers
Mac mini M6, 32GB	$1,299	7.2 tok/s estimated	Pays back in 22 years	Run the numbers
MacBook Pro M5 (14-inch), 32GB	$2,399	6.4 tok/s estimated	Pays back in 41 years	Run the numbers
Mac Studio M5 Max, 36GB	$2,499	19 tok/s estimated	Pays back in 42 years	Run the numbers
DGX Spark GB10 Grace Blackwell, 128GB	$4,699	11 tok/s estimated	Pays back in 83 years	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: 29.8B
Quantisation: Q4_K_M
Weights on disk: 17 GB
KV cache: 0.5 GB at 32k context — 3 sliding-window layers (2048) per full-attention layer, so the cache barely grows with context. Multimodal: the vision projector is a separate ~1.4 GB file, not counted here.
Maximum context: 128k tokens (128k)
Licence: Apache 2.0
Sources: source 1, source 2