Meta · First MLX build 6 Dec 2024
Llama-3.3-70B-Instruct
MLX
Text llama3.3
A 71B parameter text generation model from Meta, with 6 MLX builds, from 3-bit to bf16.
- Parameters
- 71B From Hugging Face
- Quantisation
- 4-bit Llama-3.3-70B-Instruct-4bit
- Memory needed
- 47 GB Estimate, unified memory
- Download
- 39.7 GB Weights, estimated
- Downloads
- 6.2K This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 6 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 47 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · mlx-lm$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/Llama-3.3-70B-Instruct-4bit --prompt "Hello"MLX builds
6 on mlx-community| Llama-3.3-70B-Instruct-3bit | 3-bit | 30.9 GB | 37 GB | 1.4K |
|---|---|---|---|---|
| Llama-3.3-70B-Instruct-4bit | 4-bit | 39.7 GB | 47 GB | 6.2K |
| Llama-3.3-70B-Instruct-6bit | 6-bit | 57.3 GB | 67 GB | 690 |
| Llama-3.3-70B-Instruct-8bit | 8-bit | 75.0 GB | 86 GB | 3.8K |
| llama-3.3-70b-instruct-fp16 | fp16 | 141 GB | 160 GB | 1.7K |
| Llama-3.3-70B-Instruct-bf16 | bf16 | 141 GB | 160 GB | 177 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.