Particle launches Particle-1, saying that compared with the DeepSeek original, it reduces output by approximately 45%Output Token
Particle launches a DeepSeek V4.1 post-trained on Flash,Particle-1 and says that across four benchmarks, its output Token is reduced by approximately 45%, while the task pass rate is close to the original. The model is now available at API and integrated with Vercel AI Gateway。