GPT4All Python client crashes when loading DeepSeek-R1 models

[iwr-redmond]

trafficstars

Bug Report

Attempting to load DeepSeek-R1-Distill models using the Python client fails with the error:

llama_model_load: error loading model: error loading model vocabulary: unknown pre-tokenizer type: 'deepseek-r1-qwen'

Steps to Reproduce

from gpt4all import GPT4All
model = GPT4All("DeepSeek-R1-Distill-Qwen-7B-Q4_0.gguf")

Expected Behavior

Either the GPT4All Python client should be updated to support the same models as the desktop chat application, or a separate models.json file should be used for the desktop application to ensure that users do not attempt to download and run incompatible models from the Python client.

Your Environment

• GPT4All version: 2.8.2 (python)
• Operating System: Linux
• Chat model used (if applicable): DeepSeek-R1-Distill-Qwen-7B

Notes

This bug was originally discovered by @Teriks.

[Teriks]

Updated PyPI wheels would be appreciated

I am wondering if it is technically feasible to cross compile for MacOS, or unreasonable / impossible and or non-performant