Meta (Superintelligence Lab) · Apache 2.0
Muse Glimmer 30B
Meta’s August 2026 open agent model, built to run tool-using tasks on one consumer graphics card.
How much memory it needs
The model file, plus memory for the conversation, plus about 1 GB for the app. Longer conversations need more.
| Size | Version | File | At 8K | At 32K | At 128K |
|---|---|---|---|---|---|
| 30B | Q4_K_M | 17.3 GB | 19.8 GB | 20.2 GB | 21.8 GB |
| Q8_0 | 29.6 GB | 32.1 GB | 32.5 GB | 34.1 GB |
Which computers run it
The best size for each machine at an 8K-token conversation. Speeds are estimates from memory bandwidth.
| Computer | Best version | Fit | Speed |
|---|---|---|---|
| Laptop or PC, no graphics card, 32 GB | 30B Q4_K_M | Runs well | 3–5tok/sSlow |
| RTX 4060 · 8 GB | 30B Q4_K_M | Spills into system RAM | 4–6tok/sSlow |
| RTX 3060 · 12 GB | 30B Q4_K_M | Spills into system RAM | 5–8.5tok/sSlow |
| RTX 5060 Ti · 16 GB | 30B Q4_K_M | Spills into system RAM | 8–13tok/sReadable |
| RTX 4090 · 24 GB | 30B Q4_K_M | Runs well | 35–55tok/sFast |
| RTX 5090 · 32 GB | 30B Q4_K_M | Runs great | 60–95tok/sFast |
| Mac with M4 · 16 GB | 30B Q4_K_M | Too big | — |
| Mac with M4 Pro · 48 GB | 30B Q8_0 | Runs well | 5.5–9tok/sReadable |
| Mac with M5 Max · 128 GB | 30B Q8_0 | Runs great | 12–20tok/sSmooth |
| Ryzen AI Max+ 395 · 128 GB | 30B Q4_K_M | Runs great | 9–14tok/sReadable |
| DGX Spark · 128 GB | 30B Q8_0 | Runs great | 5.5–9tok/sReadable |
| RTX PRO 6000 · 96 GB | 30B Q8_0 | Runs great | 35–55tok/sFast |
| Mac Studio with M5 Ultra · 256 GB | 30B Q8_0 | Runs great | 18–30tok/sSmooth |
| Mac Studio with M3 Ultra · 512 GB | 30B Q8_0 | Runs great | 12–20tok/sSmooth |
What it is good and bad at
Good at
- Multi-step tool use
- Reads images
- Small conversation memory
Not good at
- Image reader adds 1.4–3.9 GB on top of the model
Start it
Install Ollama, then run one of these in a terminal. LM Studio users can search the same name in the app.
ollama run muse-glimmer:30b