Can Mac M2 Max 96G run K2 Horizon MoVA 36B-A4B?
Comfortable: GGUF Q8_0 needs 42.5 GB at 4K, batch 1; headroom 29.5 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 | 24.2 | Comfortable | Verified file |
| GGUF Q5_K_M | 28.5 | Comfortable | Verified file |
| GGUF Q6_K | 33.0 | Comfortable | Verified file |
| GGUF Q8_0 | 42.5 | Comfortable | Verified file |
Context capacity · GGUF Q8_0
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 42.5 | Comfortable |
| 16K | 44.9 | Comfortable |
| 32K | 48.2 | Comfortable |
| 128K | 68.0 | Tight |
Generation speed
GGUF Q8_0: 99 tok/s · Bandwidth ceiling (not measured)