AMD Instinct MI250X 128GBBudget pick
B100 tok/s decode
~$15,000 MSRP
Can it run?
Tight fit
Using Q4_K_M in vLLM
Fit status
Tight fit
Decode
53.8 tok/s
TTFT
3601 ms
Safe context
18K
Memory
85.6 GB / 96.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 53.8 tok/s | 5237 ms | 35K |
| Chat | C | Tight fit | 53.8 tok/s | 1964 ms | 9K |
| Coding | C | Tight fit | 53.8 tok/s | 3601 ms | 18K |
| RAG | C | Tight fit | 53.8 tok/s | 6547 ms | 35K |
| Reasoning | C | Tight fit | 53.8 tok/s | 4255 ms | 18K |
How Mistral Small 4 119B (119B params) fits at each quantization level on RTX PRO 6000 Blackwell Server Edition 96GB (96.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 46.4 GB | Low | D40 |
Q3_K_S | 3 | 58.3 GB | Low | C43 |
NVFP4 | 4 | 66.6 GB | Medium | C44 |
Q4_K_MBest for your GPU | 4 | 72.6 GB | Medium | C45 |
Q5_K_M | 5 | 85.7 GB | High | C45 |
Q6_K | 6 | 97.6 GB | High | F0 |
Q8_0 | 8 | 127.3 GB | Very High | F0 |
F16 | 16 | 244.0 GB | Maximum | F0 |
huggingface-cli download mistral-small-4-119bUpgrade options
~$15,000 MSRP
~$15,000 MSRP
NVIDIA H200 141GBNVIDIA upgrade~$30,000 MSRP
Gaudi 3 128GBBiggest leap