Meta · First MLX build 23 Jul 2024
Llama-3.1-8B-Instruct
MLX
Text llama3.1
A 8B parameter text generation model from Meta, with 8 MLX builds, from 3-bit to bf16.
- Parameters
- 8B From Hugging Face
- Quantisation
- 4-bit Llama-3.1-8B-Instruct-4bit
- Memory needed
- 8 GB Estimate, unified memory
- Download
- 4.5 GB Weights, estimated
- Downloads
- 4.8K This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 8 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 8 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-3.1-8B-Instruct-4bit --prompt "Hello"MLX builds
8 on mlx-community| Meta-Llama-3.1-8B-Instruct-3bit | 3-bit | 3.5 GB | 6 GB | 99 |
|---|---|---|---|---|
| Llama-3.1-8B-Instruct-4bit | 4-bit | 4.5 GB | 8 GB | 4.8K |
| Meta-Llama-3.1-8B-Instruct-abliterated-4bit | 4-bit | 4.5 GB | 8 GB | 233 |
| Meta-Llama-3.1-8B-Instruct-abliterated-Q4-MLX | 4-bit | 4.5 GB | 8 GB | 99 |
| Meta-Llama-3.1-8B-Instruct-4bit-fakenews | 4-bit | 4.5 GB | 8 GB | 63 |
| Meta-Llama-3.1-8B-Instruct-bf16 | bf16 | 16.1 GB | 20 GB | 917 |
| Llama-3.1-8B-Instruct | Unstated | 16.1 GB | 20 GB | 763 |
| Meta-Llama-3.1-8B-Instruct-abliterated-bfloat16 | Unstated | 16.1 GB | 20 GB | 140 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.