NVIDIA GH200 96GBBudget pickC450.5 tok/s decode
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
61.2 tok/s
TTFT
3164 ms
Safe context
43K
Memory
34.4 GB / 92.2 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 61.2 tok/s | 4603 ms | 85K |
| Chat | C | Runs well | 61.2 tok/s | 1726 ms | 21K |
| Coding | C | Runs well | 61.2 tok/s | 3164 ms | 43K |
| RAG | C | Runs well | 61.2 tok/s | 5754 ms | 85K |
| Reasoning | C | Runs well | 61.2 tok/s | 3740 ms | 43K |
How Qwen3-Coder 30B A3B Instruct (30.5B params) fits at each quantization level on Mac Studio M1 Ultra 128GB (92.2 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 11.9 GB | Low | D32 |
Q3_K_S | 3 | 14.9 GB | Low | D33 |
NVFP4 | 4 | 17.1 GB | Medium | D34 |
Q4_K_M | 4 | 18.6 GB | Medium | D34 |
Q5_K_M | 5 | 22.0 GB | High | D35 |
Q6_K | 6 | 25.0 GB | High | D35 |
Q8_0 | 8 | 32.6 GB | Very High | D37 |
F16Best for your GPU | 16 | 62.5 GB | Maximum | C44 |
ollama run qwen-3-coder-30b-a3bhuggingface-cli download qwen-3-coder-30b-a3bUpgrade options
NVIDIA GH200 96GBBudget pick
NVIDIA H20 96GBBiggest leap