MacBook Pro M3 Max 64GBBudget pick
C11.2 tok/s decode
~$2,499 MSRP
Can it run?
Runs with offload
Using Q4_K_M in Ollama
Fit status
Runs with offload
Decode
28.2 tok/s
TTFT
6855 ms
Safe context
16K
Memory
31.2 GB / 32.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Very compromised (needs ~2.7 GB host RAM) | 26.4 tok/s | 10651 ms | 28K |
| Chat | C | Tight fit | 28.2 tok/s | 3739 ms | 9K |
| Coding | C | Runs with offload | 28.2 tok/s | 6855 ms | 16K |
| RAG | C | Very compromised (needs ~2.7 GB host RAM) | 26.4 tok/s | 13314 ms | 28K |
| Reasoning | C | Runs with offload | 28.2 tok/s | 8101 ms | 16K |
How Qwen3.5 35B A3B (35B params) fits at each quantization level on NVIDIA V100 32GB (32.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 13.7 GB | Low | D39 |
Q3_K_S | 3 | 17.2 GB | Low | C41 |
NVFP4 | 4 | 19.6 GB | Medium | C42 |
Q4_K_MBest for your GPU | 4 | 21.3 GB | Medium | C44 |
Q5_K_M | 5 | 25.2 GB | High | C44 |
Q6_K | 6 | 28.7 GB | High | C44 |
Q8_0 | 8 | 37.5 GB | Very High | F0 |
F16 | 16 | 71.8 GB | Maximum | F0 |
huggingface-cli download hf-lmstudio-community--qwen3-5-35b-a3b-ggufUpgrade options
~$2,499 MSRP
~$3,999 MSRP
NVIDIA A100 40GBBiggest leap~$10,000 MSRP
RTX PRO 5000 Blackwell 48GBNVIDIA upgrade