Instructions to use JANGQ-AI/MiniMax-M2.5-JANG_3M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use JANGQ-AI/MiniMax-M2.5-JANG_3M with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir MiniMax-M2.5-JANG_3M JANGQ-AI/MiniMax-M2.5-JANG_3M
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
Missing the model files?
Hello!
I think this upload is missing the model files. Your other MiniMax (2S) is quite good, so I’d be curious to try this one out.
Also, if you ever get bored, I’d appreciate an even bigger size (eg 4M or 4L - whichever would saturate my 128gb studio!).
Take care!
Not clear what are the advantages of a 3M or 4M version.
It would be nice to have other benchmarks than MMLU.
Thanks for responding. You might be right, but it'd be nice to try. This model card is set for 3M but I don't see any model files, so I was confused. Is it just a placeholder?
Anyways, I'm digging testing out your models. Very fun!
This is embarassing. I'll get this uploaded. let me know ur ram requirements/sizes and yeah i can make one specifically for u
No problem! I’m on a 128gb Mac Studio as my primary machine for inference.
No rush at all - just wanted to mention this.
4M coming, 86 gb, the difference between the 3m and the other one was barely noticeable, but this one should at least have noticeable improvement so that the size is worth it.
will you also release an abliterated version of 4M?
I just finished a 3L jang and a crack jang so i think i might take a break from the minimax models to finish up some work on my other things. we have minimax m2.7 coming out really soon so i feel like it's kinda a waste of time tbh lol