Can Mac M4 Max 36G run Ripple-65B-Preview?
Comfortable: GGUF IQ2_XXS needs 21.6 GB at 4K, batch 1; headroom 5.4 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 IQ1_M | 20.0 | Comfortable | Verified file |
| GGUF IQ2_XXS | 21.6 | Comfortable | Verified file |
| GGUF IQ2_XS | 23.8 | Tight | Verified file |
| GGUF IQ2_S | 27.2 | Tight | Verified file |
| GGUF IQ3_XXS | 29.9 | Does not fit | Verified file |
| GGUF IQ3_S | 34.5 | Does not fit | Verified file |
| GGUF IQ4_XS | 41.1 | Does not fit | Verified file |
| GGUF Q4_K_M | 41.2 | Does not fit | Verified file |
| GGUF Q5_K_M | 50.8 | Does not fit | Verified file |
| GGUF Q6_K | 60.9 | Does not fit | Verified file |
| GGUF Q8_0 | 72.6 | Does not fit | Verified file |
Context capacity · GGUF IQ2_XXS
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 21.6 | Comfortable |
| 16K | 22.4 | Comfortable |
| 32K | 23.5 | Comfortable |
| 128K | 30.1 | Does not fit |
Generation speed
GGUF IQ2_XXS: 22 tok/s · Bandwidth ceiling (not measured)