Can WeVisOne VL 4B run on the Mac M3 Ultra 256GB?
Yes. At 8K context the best fit is F16, using about 10.0 GB of the 192 GB usable on the Mac M3 Ultra 256GB. Estimated speed ceiling: ~93 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 2.9 GB Fits | 3.5 GB Fits | 6.9 GB Fits | 20.4 GB Fits | 455 |
| 3-bit (Q3_K_M) • | 3.3 GB Fits | 3.9 GB Fits | 7.3 GB Fits | 20.8 GB Fits | 365 |
| 4-bit (Q4_K_M) • | 3.8 GB Fits | 4.3 GB Fits | 7.7 GB Fits | 21.2 GB Fits | 301 |
| 5-bit (Q5_K_M) • | 4.2 GB Fits | 4.8 GB Fits | 8.1 GB Fits | 21.6 GB Fits | 259 |
| 6-bit (Q6_K) • | 4.6 GB Fits | 5.2 GB Fits | 8.6 GB Fits | 22.1 GB Fits | 226 |
| 8-bit (Q8_0) • | 5.6 GB Fits | 6.2 GB Fits | 9.6 GB Fits | 23.1 GB Fits | 174 |
| 16-bit (F16/BF16) • | 9.5 GB Fits | 10.0 GB Fits | 13.4 GB Fits | 26.9 GB Fits | 93 |
• real GGUF file size from mradermacher/WeVisOne-VL-4B-GGUF. Speed is a bandwidth ceiling (819 GB/s), not a benchmark. Methodology.
More
WeVisOne VL 4B on every GPU and Mac · Every model that runs on the Mac M3 Ultra 256GB · Apple Silicon guide · Try other settings in the calculator
Data: Hugging Face Hub (tencent/WeVisOne-VL-4B), Apple specs. Data as of 2026-11-17.