OpenAI confirms GPT-4.5 and a unified GPT-5

If you open ChatGPT today, deciding which model to pick has become a chore in itself. OpenAI boss Sam Altman admits he hates the model picker just as much as you do, and the company is finally overhauling its lineup to make AI feel unified again.

Orion lands first, then GPT-5 merges reasoning

According to Altman, OpenAI will next ship GPT-4.5, internally codenamed Orion. This will mark the company's final non-chain-of-thought model.

After that, OpenAI plans to merge its reasoning-focused o-series with the standard GPT line into a single system: GPT-5. Notably, the anticipated o3 will no longer ship as a standalone model, but will be baked directly into GPT-5.

The unified system is designed to figure out on its own whether a task needs a snap answer or prolonged thinking, while naturally handling voice, canvas, web search, and deep research.

Tiering is also changing. Free users will get unlimited chat access to GPT-5 at a standard intelligence level, subject to abuse caps. Plus and Pro subscribers will unlock progressively higher intelligence levels.

Why it matters

Nobody wants to babysit a model menu. Making the model smart enough to route its own compute is what AI was supposed to feel like from the start.

Giving free users unlimited access to standard GPT-5 is a massive flex. It puts immense pressure on rivals whose premium tiers barely match frontier performance.

That said, autonomous routing is tricky. If the system overthinks an easy question, users waste time waiting; if it cuts corners on a difficult prompt, users might miss the manual control they used to have.