dphn · First MLX build 5 Jan 2025
Dolphin3.0-Llama3.1-8B
MLX
Text llama3.1
A 8B parameter text generation model from dphn, with 6 MLX builds, from 4-bit to bf16.
- Parameters
- 8B From Hugging Face
- Quantisation
- 4-bit Dolphin3.0-Llama3.1-8B-MLX-4bit
- Memory needed
- 8 GB Estimate, unified memory
- Download
- 4.5 GB Weights, estimated
- Downloads
- 776 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 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/Dolphin3.0-Llama3.1-8B-MLX-4bit --prompt "Hello"MLX builds
6 on mlx-community| Dolphin3.0-Llama3.1-8B-MLX-4bit | 4-bit | 4.5 GB | 8 GB | 776 |
|---|---|---|---|---|
| Dolphin3.0-Llama3.1-8B-4bit | 4-bit | 4.5 GB | 8 GB | 280 |
| Dolphin3.0-Llama3.1-8B-MLX-6bit | 6-bit | 6.5 GB | 10 GB | 284 |
| Dolphin3.0-Llama3.1-8B-6bit | 6-bit | 6.5 GB | 10 GB | 43 |
| Dolphin3.0-Llama3.1-8B-8bit | 8-bit | 8.5 GB | 12 GB | 87 |
| Dolphin3.0-Llama3.1-8B-bf16 | bf16 | 16.1 GB | 20 GB | 56 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.