Can Radeon PRO W7900 48G run Muse Glimmer 30B?
Comfortable: GGUF Q4_K_M needs 17.7 GB at 4K, batch 1; headroom 30.3 GB.
Text input, FP16 KV cache, one card. Unified-memory capacity uses the GPU guide’s usable limit; negative headroom may require host offload. Speed source settings may differ from this 4K estimate.
Measured runs
No measured run for this pair. Sizing is an estimate, not a runtime test.
Quant memory at 4K
| Quant | Total GB | Verdict | Weight sizing |
|---|---|---|---|
| GGUF Q4_K_M | 17.7 | Comfortable | Verified file |
| GGUF Q5_K_M | 21.2 | Comfortable | Estimated weights |
| GGUF Q6_K | 24.6 | Comfortable | Estimated weights |
| GGUF Q8_0 | 31.1 | Comfortable | Estimated weights |
Context capacity · GGUF Q4_K_M
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 17.7 | Comfortable |
| 16K | 17.8 | Comfortable |
| 32K | 18.1 | Comfortable |
| 128K | 19.4 | Comfortable |
Generation speed
GGUF Q4_K_M: 27–36 tok/s · Derived range