Only with CPU offload

Muse Glimmer 30B at Q4_K_M needs 17.5 GB but only 5.3 GB is addressable, so about 73% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 2.8 tokens per second — usable for batch work, painful for chat.

27% on GPU 8K context Q4_K_M · 16.8 GB Apache 2.0 Released 10 Aug 2026 Vision

Meta's first open weights since Llama 4: a dense 30B distilled from Muse Spark for always-on local agents. Two KV heads keep the cache small.

Hardware requirements What hardware do I need to run Muse Glimmer 30B? →

At Q4_K_M with an 8K context: a 24 GB GPU (RTX 4090) or a 32 GB Mac (M4). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

The VRAM budget

weights 16.8 GB
Weights 16.8 GB KV cache @ 8K 0.18 GB Runtime overhead 0.6 GB Over budget 12.2 GB past 5.3 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
Q8_0 29.5 GB 30.3 GB — ~1.4 −0.1% ppl 25.0 GB over
Q6_K 22.8 GB 23.5 GB — ~1.9 −0.4% ppl 18.2 GB over
Q5_K_M 19.7 GB 20.4 GB — ~2.3 −0.8% ppl 15.1 GB over
Q4_K_M 16.8 GB 17.5 GB — ~2.8 −1.9% ppl 12.2 GB over
Q3_K_M 13.6 GB 14.3 GB — ~3.7 −5.4% ppl 9.0 GB over
Q2_K 11.6 GB 12.4 GB — ~4.6 −15% ppl 7.1 GB over

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (2048 tokens, 13 of 52 layers global), which is why its cache barely grows with context.

How to run it

terminal
$ llama-server \
    -hf meta-models/Muse-Glimmer-30B:Q4_K_M \
    -c 8192 -ngl 14

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 16.8 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on a GeForce RTX 3060 Laptop

ModelTotalTok/s est.
Qwen3 30B-A3B
30.5B MoE · similar size
Q4_K_M · 18.5 GB ~9.3 CPU offload
Qwen3 Coder 30B-A3B
30.5B MoE · similar size
Q4_K_M · 18.5 GB ~9.3 CPU offload
Granite 4.1 30B
28.9B · similar size
Q4_K_M · 18.9 GB ~2.6 CPU offload
GLM-4.7-Flash 30B-A3B
31.2B MoE · similar size
Q4_K_M · 18.6 GB ~10 CPU offload
Gemma 4 31B
31.3B · similar size
Q4_K_M · 20.2 GB — Won't fit
Nemotron 3.5 Lightning 30B-A3B
31.6B MoE · similar size
Q4_K_M · 18.4 GB ~9.9 CPU offload
Qwen3.8 27B
27.8B · similar size
Q4_K_M · 16.7 GB ~3.0 CPU offload

Every model against the GeForce RTX 3060 Laptop →

Muse Glimmer 30B on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
17.5 GB ~3.0 CPU offload
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
17.5 GB ~4.2 CPU offload
M4 · 16 GB
10.7 GB usable · 120 GB/s
17.5 GB — Won't fit
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
17.5 GB 36 Runs great
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
17.5 GB 35 Runs great
M4 · 32 GB
24.0 GB usable · 120 GB/s
17.5 GB 4.0 Runs great
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
17.5 GB 13 Runs great

All 114 devices, by tier →