How to use DeepSeek V3–0324 for free?
So DeepSeek, in a surprise move, has released a new model, DeepSeek V3–0324. And it’s again completely open-sourced!
https://medium.com/media/3f79923c3fb91fd47ce8a705c3830cca/href
Surprisingly, no other information is published
What is DeepSeek V3–0324?
DeepSeek V3–0324 is an updated checkpoint for DeepSeek’s old model, DeepSeek V3, which was released Dec’24. Though, nothing has been mentioned about the model features so far, early discussion on Reddit suggests some strong capabilities:
https://medium.com/media/144e145553e8f033954c6f8d4ea1804f/href
The model boasts 685B parameters and a MoE model
The model is hosted on HuggingFace (open-sourced).
The coding abilities are looking better compared to previous DeepSeek models
The checkpoint was released on March 24, 2024, hence the “0324” suffix.
The base for R2 (Reasoning Model)?
Maybe yes. Many speculate that this updated V3–0324 will serve as the foundation for DeepSeek-R2, an improved reasoning-focused model expected in April/May 2024.
No Benchmarks
- No official benchmarks have been released yet, but independent tests are expected soon.
- One user noted it effectively passes complex reasoning tasks (e.g., “bouncing ball in rotating shape”).
Tone & Personality Changes compared to DeepSeek-v3
- A few users felt the new version sounded “more robotic” compared to the original V3, which had a more human-like, conversational tone.
- Some mentioned it now feels “too intellectual” and less engaging for casual chat.
How to use DeepSeek V3–0324 for free?
As notified on Reddit, the model is updated on
Also, the weights are available here
deepseek-ai/DeepSeek-V3-0324 · Hugging Face
Will be updating the blog once more info is available. By the time, try the new model by DeepSeek!
DeepSeek V3-0324: New DeepSeek model released was originally published in Data Science in your pocket on Medium, where people are continuing the conversation by highlighting and responding to this story.
