DeepSeek-R1-Distill-Qwen-7B: Copyright (c) 2023 DeepSeek. MIT License.
https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B

This checkpoint is derived from Alibaba's Qwen 2.5 series (Apache-2.0), not
from a Llama model. Both DeepSeek's MIT text and Qwen's Apache-2.0 text are
included to retain upstream rights and notices.

mlx-community converted it to 4-bit MLX with mlx-lm 0.21.1:
https://huggingface.co/mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit

Noted changes the chat template to omit the initial thinking marker and to
accept a system prompt. It does not modify the model weights. The legacy
preference ID deepseek-r1-8b selects this Qwen-based 7B checkpoint.