OpenAI is retiring GPT-4.5 — its largest and most compute-heavy model to date — from its API offering, just months after its February launch. The company announced on Monday that API access to GPT-4.5 will end on July 14, encouraging developers to transition to the newly launched GPT-4.1 instead.
A Short Run for a Giant Model
Despite being touted as OpenAI’s most powerful model when it debuted, GPT-4.5 (code-named Orion) is being phased out due to high operational costs and the rise of leaner, more efficient alternatives. GPT-4.1, which also launched Monday, is being positioned as a direct upgrade.
“[GPT-4.1] offers similar or improved performance than GPT-4.5 in key areas at a much lower cost,” an OpenAI spokesperson told TechCrunch, adding that retiring GPT-4.5 allows the company to focus on “building future models.”
Still in ChatGPT — Just Not for API Users
GPT-4.5 isn’t disappearing entirely — it’ll remain available within ChatGPT’s research preview for Plus subscribers. But for developers relying on OpenAI’s API, the clock is ticking.
Why the Exit?
While GPT-4.5 delivered strong results in writing and reasoning tasks, it fell short of hitting "frontier" status across industry benchmarks. The model was also extremely expensive to operate, with OpenAI previously cautioning that it might not be sustainable for API use.
Pricing for GPT-4.5 has reflected those challenges: $75 per million input tokens and $150 per million output tokens, making it one of the priciest models in OpenAI’s lineup.
What’s Next?
With GPT-4.1 now on the scene — offering better value, similar performance, and easier scalability — OpenAI is betting developers will make the switch quickly. As the AI arms race heats up, the company appears to be streamlining its model offerings to focus on what’s next.
AI Agents