OpenAI plans to launch Orion, its next major AI model, in December.

Unlike previous releases, Orion won’t be widely available on ChatGPT right away. Instead, OpenAI will first give early access to specific partner companies, so they can create their own products using the new model.

Microsoft, OpenAI’s main partner, is getting ready to host Orion on Azure, possibly by November.

Within OpenAI, Orion is seen as a natural step after GPT-4, though it’s not certain it will be called GPT-5.

In brief:

• Orion will be released first to specific partners, not directly on ChatGPT.
• Microsoft is preparing to support Orion on Azure, potentially by November.
• Orion could be up to 100 times stronger than GPT-4, moving toward AGI goals.

VIPs only

Orion’s training used data from OpenAI’s o1 model, codenamed “Strawberry”, finishing in September.

OpenAI’s CEO Sam Altman hinted at Orion’s release with a mention of “winter constellations”.

This launch also aligns with OpenAI’s new for-profit direction after training $6.6 billion in funding, but key leaders, including CTO Mira Murati, recently left the company during the transition.

Winter’s coming—and so is a mega-powerful AI model.