OpenAI’s next-gen AI model, Orion, coming sooner than expected, with 100x the power of GPT-4: Report

tech.hindustantimes.com:

OpenAI, the Microsoft-backed creator of ChatGPT, could launch its new AI model, called Orion, in December. However, the company plans to approach this launch differently, making it initially available only to its partner companies. It may not be accessible in ChatGPT, as was the case with the last two models—GPT-4o and GPT-o1 “Strawberry”—according to The Verge. The publication reports that OpenAI will allow partner companies like Microsoft to use the AI models to build their own products and additional features.

Also Read: OpenAI's next-gen AI model, Orion, coming sooner than expected, with 100x the power of GPT-4

The report adds that Microsoft, one of OpenAI's major investors, is planning to integrate the Orion AI model into its Azure services as soon as November. However, it remains unclear if OpenAI will continue using the name Orion publicly, or if it will ultimately be branded as GPT-5, which is more widely recognised.

For context, GPT-4 was launched in March 2023, and since then we have seen the company launch other models, including GPT-4o, and more recently, GPT-o1, codenamed “Strawberry.” 

Also Read: Apple October event confirmed: M4 Macs launching next week

Orion, or whatever OpenAI ultimately names it externally, is said to be 100 times more powerful than GPT-4. However, it should not be confused with models like GPT-o1 “Strawberry” or GPT-4o. The report suggests that Orion may have been trained using data from OpenAI o1, enhancing its advanced reasoning capabilities and ability to process larger requests. That said, OpenAI's Orion is likely to remain a closed model, meaning it won't be openly available like Meta's Llama family of AI models.

Also Read: ‘Important that we purvey accurate information, not fantasy': Apple on why iPhone lacks Google Pixel-like AI