Can it run?
Runs well
Using Q4_K_M in Ollama
Fit status
Runs well
Decode
128.1 tok/s
TTFT
1511 ms
Safe context
16K
Memory
7.6 GB / 10.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 135.3 tok/s | 2082 ms | 16K |
| Chat | B | Runs well | 135.3 tok/s | 781 ms | 11K |
| Coding | B | Runs well | 128.1 tok/s | 1511 ms | 16K |
| RAG | C | Tight fit | 135.3 tok/s | 2602 ms | 16K |
| Reasoning | B | Runs well | 135.3 tok/s | 1691 ms | 16K |
How CodeLlama 7B Instruct (7B params) fits at each quantization level on RTX 3080 10GB (10.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 2.7 GB | Low | D34 |
Q3_K_S | 3 | 3.4 GB | Low | D36 |
NVFP4 | 4 |
ollama run codellama-7b-instructhuggingface-cli download codellama-7b-instruct| Medium |
| D37 |
Q4_K_M | 4 | 4.3 GB | Medium | D38 |
Q5_K_M | 5 | 5.0 GB | High | D39 |
Q6_KBest for your GPU | 6 | 5.7 GB | High | C41 |
Q8_0 | 8 | 7.5 GB | Very High | C43 |
F16 | 16 | 14.3 GB | Maximum | F0 |