OpenAI ships o1 and a $200-a-month tier

OpenAI opened its “12 Days of Shipmas” run of announcements by releasing the full version of o1, its reasoning model, to ChatGPT and introducing ChatGPT Pro, a $200-a-month subscription tier. The full o1 replaced o1-preview, which had been available since September, and OpenAI said the finished model made major mistakes roughly a third less often while responding faster, and could now reason over uploaded images as well as text.

ChatGPT Pro’s headline feature was unlimited use of o1, o1-mini, GPT-4o and Advanced Voice Mode, alongside “o1 pro mode,” a variant that applied more compute per query to give more reliable answers on harder problems. OpenAI pitched the tier at researchers, engineers and other users who relied on the model daily rather than casual chatbot users, positioning it well above the existing $20-a-month Plus subscription and closer to enterprise software pricing than any previous consumer AI product.

The jump from $20 to $200 a month drew attention less for what it added than for what it signalled: that OpenAI believed some users would pay a premium simply for more inference-time compute applied to their queries, a bet that reasoning models’ extended “thinking” time was itself a saleable resource rather than an engineering cost to be hidden inside a flat subscription. It set a template — tiered access gated by how much compute a query could consume — that other labs and OpenAI’s own later releases continued to build on.

The release came two days before the start of a run of daily OpenAI announcements through mid-December, of which the $200 tier and full o1 were the opening act; o1 reached the API with developer tooling later the same month.