Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
436.4 tok/s
TTFT
444 ms
Safe context
96K
Memory
30.9 GB / 184.3 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 436.4 tok/s | 645 ms | 191K |
| Chat | C | Runs well | 436.4 tok/s | 350 ms | 48K |
| Coding | C | Runs well | 436.4 tok/s | 444 ms | 96K |
| RAG | C | Runs well | 436.4 tok/s | 807 ms | 191K |
| Reasoning | C | Runs well | 436.4 tok/s | 524 ms | 96K |
How Gemmasutra Mini 2B v1 (2B params) fits at each quantization level on Mac Studio M3 Ultra 256GB (184.3 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 0.8 GB | Low | D29 |
Q3_K_S | 3 | 1.0 GB | Low | D30 |
NVFP4 | 4 |
1.1 GB |
| Medium |
| D30 |
Q4_K_M | 4 | 1.2 GB | Medium | D30 |
Q5_K_M | 5 | 1.4 GB | High | D30 |
Q6_K | 6 | 1.6 GB | High | D30 |
Q8_0 | 8 | 2.1 GB | Very High | D30 |
F16Best for your GPU | 16 | 4.1 GB | Maximum | D30 |
