From www.tomshardware.com
OpenAI is set to introduce Orion, its next-generation AI model, this December, reports The Verge, citing its sources with knowledge of the matter. However, initial access will be limited to key partner companies instead of a broad release through ChatGPT.com to the general public. The new model is expected to be a full-blown version rather than an enhanced or specialized version of an existing one.
Orion is viewed internally as a successor to GPT-4, though it is unclear whether its official name will be GPT-5 when released. An OpenAI executive has reportedly hinted that Orion could be up to 100 times more powerful than GPT-4, Open AI’s flagship model. Unlike the recently released o1 model, designed for enhanced reasoning and aimed at applications like scientific problem-solving, data analysis, and structured knowledge tasks, Orion is geared toward broader AI capabilities for the general public.
For now, OpenAI plans to keep AI models with enhanced reasoning and AI models with broader capabilities separate, though eventually, they will merge as the company is working on artificial general intelligence (AGI). Yet, Orion’s training involved synthetic data generated by o1, known internally as Strawberry.
According to The Verge, engineers at Microsoft Azure, OpenAI’s cloud service provider, are getting ready to launch Orion on the Azure platform, potentially starting in November.
Earlier this year, a source informed The Verge that in September, OpenAI researchers organized a happy hour event to celebrate the new model’s completion of the training phase. Around the same time, Sam Altman, chief executive of OpenAI, posted an X message about winter constellation in the U.S. Midwest. Curiously, The Verge tried to ask the o1 AI model to decipher what Sam Altman’s post means, and it came up with the answer that it hides the word ‘Orion.’ Yet, when yours truly asked to do the same, it responded that the message hides the phrase ‘It Begins.’ Nonetheless, all AI models responded that the post could hint at a big announcement.
The timing of Orion’s release is pivotal for OpenAI, coinciding with the organization’s transition to a for-profit entity. Perhaps this is why the company focuses on revealing it to partners rather than the general public first. This shift comes from the recent funding round that raised $6.6 billion.
[ For more curated tech news, check out the main news page here]
The post OpenAI will reportedly unleash next-gen Orion AI model this December — Orion is expected to be 100X more potent than GPT-4 first appeared on www.tomshardware.com