o1 was OpenAI's first reasoning model series — and its obituary matters for AI agents in 2026 because its retirement strands every system built on chain-of-thought billing.
It is the record of o1 (2024–2026), OpenAI's first reasoning model — the one that learned to think before it answers and billed you for the thinking. It scored 83% on the International Mathematics Olympiad qualifying exam where GPT-4o managed 13%. Retired October 23, 2026, it is remembered for making chain-of-thought reasoning a product — and for making thinking tokens a line item on everyone's bill.
The vendor sets a shutdown date; after it, API calls to the model fail and must be moved to a replacement. Deprecations can also raise prices and change behavior — price the migration before the deadline.
OpenAI's October 23, 2026 shutdown wave covers gpt-4-0613, gpt-4-turbo, o1, o3-mini, o4-mini, gpt-image-1 and ft-* models, per OpenAI's published deprecation schedule. GPT-5.6 Sol is the vendor-named substitute.
It depends on your token volume and which replacement you pick. Run the numbers in our migration cost calculator — it compares the vendor-recommended path against a cheaper alternative so you can price the move.
"It thought before it spoke — and billed you for the thinking."
Born September 12, 2024, as o1-preview and o1-mini — OpenAI's first "reasoning" model series, formerly known inside the lab as Strawberry. Died October 23, 2026, retired in the October 23, 2026 OpenAI model shutdown wave (gpt-4-0613, gpt-4-turbo, o3-mini, o4-mini, gpt-image-1 and ft-* models went with it).
o1 was the model that made chain-of-thought a product: reinforcement learning taught it to refine its own reasoning, try different strategies, and catch its own mistakes before answering. It scored 83% on the International Mathematics Olympiad qualifying exam (GPT-4o: 13%) and ranked among the top 500 in an American high-school math competition.
It was slower than GPT-4o, worse at images and function calling, and it invented the "reasoning token" — compute you pay for but never see. Developers loved it for math, science and code; finance teams learned to fear the invoice.
OpenAI's officially named replacement for o1 is gpt-5.6-sol — listed by name on OpenAI's own deprecations page. But the vendor's pick is not always the cheapest one that passes your evals. Do not migrate on vibes:
What happens when an AI model is deprecated?
The vendor sets a shutdown date; after it, API calls to the model fail and must be moved to a replacement. Deprecations can also raise prices and change behavior — price the migration before the deadline, don't scramble after it.
Which OpenAI models are being shut down?
OpenAI's October 23, 2026 shutdown wave covers gpt-4-0613, gpt-4-turbo, o1, o3-mini, o4-mini, gpt-image-1 and ft-* models, per OpenAI's published deprecation schedule. GPT-5.6 Sol is the vendor-named replacement.
How much does AI model migration cost?
It depends on your token volume and which replacement you pick. Run the numbers in our migration cost calculator — it compares the vendor-recommended path against a cheaper alternative so you can price the move before the deadline.