AMD Instinct MI300X 192GBBudget pick
C256.7 tok/s decode
~$15,000 MSRP
Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
34.6 tok/s
TTFT
5598 ms
Safe context
38K
Memory
78.4 GB / 184.3 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 34.6 tok/s | 8143 ms | 75K |
| Chat | C | Runs well | 34.6 tok/s | 3054 ms | 19K |
| Coding | C | Runs well | 34.6 tok/s | 5598 ms | 38K |
| RAG | C | Runs well | 34.6 tok/s | 10178 ms | 75K |
| Reasoning | C | Runs well | 34.6 tok/s | 6616 ms | 38K |
How Qwen3-Coder-Next (80B params) fits at each quantization level on Mac Studio M3 Ultra 256GB (184.3 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 31.2 GB | Low | D33 |
Q3_K_S | 3 | 39.2 GB | Low | D34 |
NVFP4 | 4 | 44.8 GB | Medium | D35 |
Q4_K_M | 4 | 48.8 GB | Medium | D35 |
Q5_K_M | 5 | 57.6 GB | High | D36 |
Q6_K | 6 | 65.6 GB | High | D37 |
Q8_0Best for your GPU | 8 | 85.6 GB | Very High | D40 |
F16 | 16 | 164.0 GB | Maximum | C45 |
ollama run qwen-3-coder-nexthuggingface-cli download qwen-3-coder-nextUpgrade options
~$15,000 MSRP
NVIDIA GB200 192GBBest value~$60,000 MSRP
H100 NVL 188GBBiggest leap