OpenAI releases o3-mini, including for free ChatGPT users

Background

OpenAI announced o3-mini on 20 December 2024, together with a larger model called o3^[1]. The name placed it in the company's reasoning series: models that work through intermediate steps before producing a final answer, spending extra computation on a problem instead of responding immediately. Until the end of January 2025, that kind of model was something paying customers could reach; free users of ChatGPT could not.

The release

On 31 January 2025 OpenAI released o3-mini to all ChatGPT users, including free users, and to some users of its API^[1]. The company described it as the newest and most cost-efficient model in its reasoning series, and said it was available in ChatGPT and in the API from launch^[3].

How it was offered

Free users received the medium reasoning-effort version of the model, while paying users could also choose o3-mini-high^[1]. "Reasoning effort" is a setting for how much work the model is allowed to do before answering: a higher setting gives it more room to think a problem through, and a lower one answers faster and cheaper. Offering two levels let the same model be sold at different points on that trade-off.

Inside ChatGPT the model appeared as a "Reason" option in the message composer, the box where a user types a prompt. It was the first reasoning model offered to free ChatGPT users, who could pick that option themselves^[3].

The launch also came with a change to usage limits. OpenAI tripled the daily rate limit for Plus and Team subscribers, from 50 to 150 messages per day^[3]. Enterprise customers were scheduled to gain access in February 2025^[3].

For developers, the API was priced at $1.10 per million input tokens and $4.40 per million output tokens^[2]. Tokens are the small chunks of text a model reads and writes, and quoting API prices per million of them is the usual way such access is sold.

Why it matters

The release is a marker of how access to a new class of model widened. A reasoning model had been a paid-tier feature; with o3-mini it became part of the free tier of ChatGPT, selectable through a single option in the interface^[3]. At the same time the model was positioned on cost: OpenAI called it the most cost-efficient in the series^[3], and the API price list set out what that meant in practice for developers paying per token^[2].

The accompanying jump in the message limit for Plus and Team users, from 50 to 150 per day, shows that the constraint on using these models was not only capability but how often a subscriber could call on them^[3]. Enterprise access was handled on a separate schedule, set for the following month^[3].

Read ten years on, the event is a clear data point for when reasoning models stopped being a premium extra and became something an ordinary ChatGPT user could try: announced in December 2024^[1], shipped at the end of January 2025 to everyone including free users^[1], with a cheaper per-token API price^[2] and a higher message allowance for subscribers^[3].