Falcon 40B — 40B Dense.
Run Falcon 40B locally with Ollama — needs 24.9 GB VRAM at Q4_K_M:
ollama run falcon:40b| Quant | Bits | VRAM @ 8K | Quality |
|---|---|---|---|
| IQ3_XXS | 3.25 | 17.2 GB 16.7 + 0.5 KV | low |
| IQ3_XS | 3.5 | 18.5 GB 18.0 + 0.5 KV | low |
| Q3_K_S | 3.64 | 19.2 GB 18.7 + 0.5 KV | low |
| IQ3_M | 3.76 | 19.8 GB 19.3 + 0.5 KV | low |
| Q3_K_M | 4 | 21.0 GB 20.5 + 0.5 KV | low |
| Q3_K_L | 4.3 | 22.5 GB 22.0 + 0.5 KV | moderate |
| IQ4_XS | 4.46 | 23.3 GB 22.8 + 0.5 KV | moderate |
| Q4_K_S | 4.67 | 24.3 GB 23.8 + 0.5 KV | moderate |
| Q4_K_M | 4.89 | 25.4 GB 24.9 + 0.5 KV | good |
| Q5_K_S | 5.57 | 28.8 GB 28.3 + 0.5 KV | good |
| Q5_K_M | 5.7 | 29.5 GB 29.0 + 0.5 KV | good |
| Q6_K | 6.56 | 33.8 GB 33.3 + 0.5 KV | excellent |
| Q8_0 | 8.5 | 43.5 GB 43.0 + 0.5 KV | lossless |
| FP16 | 16 | 81.0 GB 80.5 + 0.5 KV | lossless |
Select your GPU above to see speed estimates and compatibility for each quantization.
Rent a GPU and run Falcon 40B now
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for Falcon 40B
Build Hardware for Falcon 40BFalcon 40B — was the top open model in mid-2023. Still decent for its size.
Run Falcon 40B locally with Ollama — needs 24.9 GB VRAM at Q4_K_M:
ollama run falcon:40b