Can it run?
Runs well
Using Q5_K_M in llama.cpp
Fit status
Runs well
Decode
79.3 tok/s
TTFT
2441 ms
Safe context
8K
Memory
22.4 GB / 40.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 92.5 tok/s | 3044 ms | 8K |
| Chat | C | Runs well | 92.5 tok/s | 1141 ms | 8K |
| Coding | C | Runs well | 79.3 tok/s | 2441 ms | 8K |
| RAG | B | Runs well | 92.5 tok/s | 3804 ms | 8K |
| Reasoning | C | Runs well | 92.5 tok/s | 2473 ms | 8K |
How InternLM 20B (20B params) fits at each quantization level on NVIDIA A100 40GB (40.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 7.8 GB | Low | D33 |
Q3_K_S | 3 | 9.8 GB | Low | D34 |
NVFP4 | 4 |
huggingface-cli download internlm-20b11.2 GB |
| Medium |
| D35 |
Q4_K_M | 4 | 12.2 GB | Medium | D36 |
Q5_K_M | 5 | 14.4 GB | High | D37 |
Q6_K | 6 | 16.4 GB | High | D38 |
Q8_0Best for your GPU | 8 | 21.4 GB | Very High | C41 |
F16 | 16 | 41.0 GB | Maximum | F0 |