DeepSeek, which showcased low-cost and high-performance generative artificial intelligence (AI), has decided to unveil its next inference model earlier than planned.
Reuters reported on the 25th (local time) that DeepSeek plans to launch its next inference model R2 earlier than the initially planned early May release. DeepSeek expects the new model R2 to generate improved coding and to be capable of inference in languages other than English.
DeepSeek introduced the large language model (LLM) V3 last December and made R1 open source on the 20th of last month. R1 was developed at less than one-tenth the expense of the top U.S. models, demonstrating a performance level similar to OpenAI's ChatGPT, which shocked the U.S. tech industry.
Meanwhile, OpenAI also announced on the 12th that it will launch the next-generation model "GPT-4.5." Sam Altman, CEO of OpenAI, noted through social media that "this will be the last non-cognitive chain model we develop."