Hacker News

17 points by meetpateltech 1 day ago | 7 comments

From what my AIs tell me, Particle-1 is post-trained on DeepSeek V4.1 Flash 1017 which is important. It's also said to be open weight and I see Particle AI[0] on Hugging Face but no published models.

[0] https://huggingface.co/ParticleAI

[–] nateb2022 17 hours ago

> From what my AIs tell me, Particle-1 is post-trained on DeepSeek V4.1 Flash 1017 which is important.

If you clicked the submission, that's the first sentence of the second paragraph. You would probably have saved time doing that than consulting your AIs.

[–] rvz 1 day ago

> From what my AIs tell me, Particle-1 is post-trained on DeepSeek V4.1 Flash 1017

So then what is the point of this model that I could not already do with DeepSeek?

Why does "Particle" even exist?

[–] stingrae 1 day ago

That is sort of like saying why apply reinforcement learning (most of the modern improvements in LLMs have come from RL) to any raw LLM. They are refining it to perform better at certain tasks.

[–] rvz 22 hours ago

> They are refining it to perform better at certain tasks.

Like what, coding?

Over 50+ LLMs out there do exactly this.

[–] kadoban 1 day ago

Can't tell if they plan to release the model weights? Very uninteresting if not. Otherwise seems like potentially some nice post training.

[–] mono442 1 day ago

A terrible slopped website. They could have put a bit more effort into an actual design.