Autor: Jakub Rusinowski · Ostatnia aktualizacja: 8 grudnia 2024
Najnowszy flagowy model Meta. Llama 3.3 70B dorównuje w wielu benchmarkach ogromnemu modelowi 405B, pozostając wystarczająco wydajnym do wysokiej klasy lokalnych konfiguracji.
| Licence | What it permits | Applies to |
|---|---|---|
Llama Community | Commercial use permitted Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms. | Llama 3.3 70B Instruct |
| Llama 3.3 70B Instruct | Min 43 GB VRAM · Q2_K_XS (Tight) · 128,000 ctx · ollama run llama3.3 |
The cheapest GPU that runs Llama 3.3 locally (min 43 GB VRAM) is the Apple M5 Pro (64 GB).
Install Ollama then run: ollama run llama3.3
Minimum VRAM: 43 GB. For best results use Q4_K_M quantization.
Llama 3.3 needs about 43 GB VRAM at Q2_K_XS (Tight) quantization for its smallest variant. Variants: Llama 3.3 70B Instruct (43 GB, Q2_K_XS (Tight)). On Apple Silicon, unified memory counts toward this requirement.
Llama 3.3's smallest variant needs about 43 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
Q4_K_M is the best balance of quality and VRAM for Llama 3.3 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 llama3.3. This downloads Llama 3.3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.