Can it run?
Runs well
Using Q6_K in Ollama
Fit status
Runs well
Decode
136.5 tok/s
TTFT
1418 ms
Safe context
63K
Memory
5.9 GB / 23.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Runs well | 147.0 tok/s | 1916 ms | 126K |
| Chat | C | Runs well | 147.0 tok/s | 718 ms | 31K |
| Coding | C | Runs well | 136.5 tok/s | 1418 ms | 63K |
| RAG | C | Runs well | 147.0 tok/s | 2395 ms | 126K |
| Reasoning | C | Runs well | 147.0 tok/s | 1556 ms | 63K |
How SmolVLM 500M Instruct (0.5B params) fits at each quantization level on MacBook Pro M2 Pro 32GB (23.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 0.2 GB | Low | D30 |
Q3_K_S | 3 | 0.2 GB | Low | D30 |
NVFP4 | 4 |
huggingface-cli download hf-ggml-org--smolvlm-500m-instruct-gguf0.3 GB |
| Medium |
| D30 |
Q4_K_M | 4 | 0.3 GB | Medium | D30 |
Q5_K_M | 5 | 0.4 GB | High | D30 |
Q6_K | 6 | 0.4 GB | High | D30 |
Q8_0 | 8 | 0.5 GB | Very High | D30 |
F16Best for your GPU | 16 | 1.0 GB | Maximum | D30 |