Yes — with 60.9 GB to spare

Muse Glimmer 30B at Q4_K_M fits your H100 PCIe entirely on the GPU at 8K context, at an estimated 72 tokens per second. There is room for its full 128K window.

Fully on GPU 8K context Q4_K_M · 16.8 GB Apache 2.0 Released 10 Aug 2026 New this month 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.

What hardware do I need for Muse Glimmer 30B? →

The VRAM budget

weights 16.8 GB

Weights 16.8 GB KV cache @ 8K 0.18 GB Runtime overhead 0.6 GB Free 60.9 GB of 78.4 GB

Quantisation ladder

Quant	Weights	Total @ 8K	Max context	Tok/s	Quality	Fit
Q8_0	29.5 GB	30.3 GB	128K	41	−0.1% ppl	Long context
Q6_K	22.8 GB	23.5 GB	128K	53	−0.4% ppl	Long context
Q5_K_M	19.7 GB	20.4 GB	128K	62	−0.8% ppl	Long context
Q4_K_M	16.8 GB	17.5 GB	128K	72	−1.9% ppl	Recommended
Q3_K_M	13.6 GB	14.3 GB	128K	89	−5.4% ppl	Long context
Q2_K	11.6 GB	12.4 GB	128K	104	−15% ppl	Long context

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

Ollama llama.cpp LM Studio vLLM

terminal

$ ollama pull muse-glimmer:30b
$ OLLAMA_CONTEXT_LENGTH=8192 \
    ollama run muse-glimmer:30b

The default. One binary, a model registry, an OpenAI-compatible port. More on Ollama.

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.

03There is room to go to the model's full 128K context on this card.

See all models for this rig Compare with another model