Autor: Jakub Rusinowski · Ostatnia aktualizacja: 18 grudnia 2024
Najnowsza generacja modeli Technology Innovation Institute, wytrenowanych na 14 bilionach tokenów. Falcon 3 osiąga wyniki SOTA dla swojej liczby parametrów i jest wydawany na w pełni permisywnej licencji Apache 2.0 — jeden z najlepszych wyborów do wdrożeń komercyjnych.
| Licence | What it permits | Applies to |
|---|---|---|
Apache-2.0 | Commercial use permitted Commercial use permitted. No usage restrictions beyond attribution. | Falcon 3 3B Instruct, Falcon 3 7B Instruct, Falcon 3 10B Instruct |
| Falcon 3 3B Instruct | Min 3 GB VRAM · Q4_K_M · 32,768 ctx · ollama run falcon3:3b |
| Falcon 3 7B Instruct | Min 5 GB VRAM · Q4_K_M · 32,768 ctx · ollama run falcon3:7b |
| Falcon 3 10B Instruct | Min 7 GB VRAM · Q4_K_M · 32,768 ctx · ollama run falcon3:10b |
The cheapest GPU that runs Falcon 3 locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run falcon3:3b
Minimum VRAM: 3 GB. For best results use Q4_K_M quantization.
Falcon 3 needs about 3 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Falcon 3 3B Instruct (3 GB, Q4_K_M); Falcon 3 7B Instruct (5 GB, Q4_K_M); Falcon 3 10B Instruct (7 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Yes — Falcon 3 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
Q4_K_M is the best balance of quality and VRAM for Falcon 3 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run falcon3:3b. This downloads Falcon 3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.