OpenAI has successfully concluded the pre-training phase of its new AI model, codenamed Bel, which boasts more than 10 trillion parameters. This substantial model is designed to serve as the foundation for the upcoming GPT-6 and subsequent AI iterations beyond it.
Bel is anticipated to be a significant advancement in AI, potentially positioning itself as the first model approaching artificial general intelligence (AGI). The scale of Bel's pre-training aligns it closely with the capabilities of GPT-4.5, indicating a leap forward in AI technology.
The completion of Bel's pre-training raises concerns at competitor Anthropic, which is reportedly struggling to mount an adequate response to OpenAI's advancements. Sources suggest that Anthropic is hampered by limited computational resources and may spend the remainder of the year at a disadvantage compared to OpenAI. However, they are optimistic about regaining their competitive edge by early next year.
While the launch of GPT-6 is still pending, it will take some time before solutions derived from the Bel model become available to the public. This development underscores OpenAI's lead in the AI market, increasing pressure on competitors to innovate rapidly in order to keep pace.
Informational material. 18+.
