Yes — with 54.5 GB to spare
Muse Glimmer 30B at Q4_K_M fits your M2 Max · 96 GB entirely in unified memory at 8K context, at an estimated 13 tokens per second. There is room for its full 128K window.
Fully in unified memory
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.
The VRAM budget
weights 16.8 GB
Weights 16.8 GB
KV cache @ 8K 0.18 GB
Runtime overhead 0.6 GB
Free 54.5 GB of 72.0 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 29.5 GB | 30.3 GB | 128K | 7.6 | −0.1% ppl | Long context |
| Q6_K | 22.8 GB | 23.5 GB | 128K | 9.8 | −0.4% ppl | Long context |
| Q5_K_M | 19.7 GB | 20.4 GB | 128K | 11 | −0.8% ppl | Long context |
| Q4_K_M | 16.8 GB | 17.5 GB | 128K | 13 | −1.9% ppl | Recommended |
| Q3_K_M | 13.6 GB | 14.3 GB | 128K | 16 | −5.4% ppl | Long context |
| Q2_K | 11.6 GB | 12.4 GB | 128K | 19 | −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
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/Muse-Glimmer-30B-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 16.8 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02macOS caps what the GPU may wire down at about 72.0 GB of 96 GB. Raising it with iogpu.wired_limit_mb is possible, and risky.
03There is room to go to the model's full 128K context on this card.