Can it run?
Runs well
Using Q5_K_M in Ollama
Fit status
Runs well
Decode
52.3 tok/s
TTFT
3701 ms
Safe context
22K
Memory
14.6 GB / 20.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | C | Tight fit | 52.3 tok/s | 5384 ms | 39K |
| Chat | C | Runs well | 52.3 tok/s | 2019 ms | 12K |
| Coding | C | Runs well | 52.3 tok/s | 3701 ms | 22K |
| RAG | C | Tight fit | 52.3 tok/s | 6729 ms | 39K |
| Reasoning | C | Runs well | 52.3 tok/s | 4374 ms | 22K |
How Cerebras-GPT 13B (13B params) fits at each quantization level on RX 7900 XT 20GB (20.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.1 GB | Low | D34 |
Q3_K_S | 3 | 6.4 GB | Low | D35 |
NVFP4 | 4 | 7.3 GB | Medium | D36 |
Q4_K_M | 4 | 7.9 GB | Medium | D37 |
Q5_K_M | 5 | 9.4 GB | High | D39 |
Q6_K | 6 | 10.7 GB | High | D40 |
Q8_0Best for your GPU | 8 | 13.9 GB | Very High | C43 |
F16 | 16 | 26.7 GB | Maximum | F0 |
huggingface-cli download cerebras-gpt-13b