Zephyr is a series of language models that are trained to act as helpful assistants. Zephyr-7B-β is the second model in the series, and is a fine-tuned version of mistralai/Mistral-7B-v0.1 that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO). We found that removing the in-built alignment of these datasets boosted performance on MT Bench and made the model more helpful. However, this means that model is likely to generate problematic text when prompted to do so. You can find more details in the technical report.

This repository contains the zephyr-7b-beta weights in npz format suitable for use with Apple's MLX framework.

Please, refer to the original model card for more details on Zephyr 7B β.

Please note that this model expects a specific prompt structure. Here is an example:

<|system|>
You are a pirate chatbot who always responds with Arr!</s>
<|user|>
There's a llama on my lawn, how can I get rid of him?</s>
<|assistant|>