O3-Mini Launching Soon: Sam Altman Claims AGI Needs Only 872 MW of Compute Power
Explore the latest on OpenAI's o3-mini launch, AI model distillation, and Sam Altman's insights on AGI's 872 MW power requirements. Don't miss out!
There’s been speculation in the industry that leading companies like OpenAI and Anthropic may have already trained their next-generation large models. However, due to their high usage costs, these models are unlikely to be released in the short term.
To meet user expectations, companies like OpenAI and Anthropic are releasing smaller distilled models based on their next-generation large models.
OpenAI CEO Sam Altman announced on 𝕏 that the o3-mini model will be launched "within a few weeks," with both API and web-based versions available simultaneously.
The name itself suggests that the o3-mini is a distilled version of a larger model, supporting the above speculation.
According to OpenAI research scientist Hongyu Ren, three versions of o3-mini—high, medium, and low—will be released simultaneously.
This news isn't entirely new, as on the last day of OpenAI's 12-day streak of announcements last month, Sam Altman had already mentioned plans to launch o3-mini around the end of January, followed by the full o3 model. For more details, you can visit this article.
After the announcement, Altman actively responded to user questions, revealing more details.