Sam Altman says o3-mini is coming to free ChatGPT users

OpenAI usually reserves its shiny new models for paying subscribers, leaving free users stuck with last-generation tech. Sam Altman just flipped that playbook.

On Thursday, January 23, the OpenAI CEO confirmed on X that o3-mini, the lightweight version of the company’s next-generation reasoning family, is heading to free ChatGPT users as well. Paid subscribers on Plus, Teams, and Pro plans will still get priority, with Altman promising Plus members tons of usage.

This follows Altman’s announcement earlier in the week that the full o3 model had passed external safety testing and is set to arrive across the API and ChatGPT in the coming weeks. The flagship model was first teased in December during OpenAI’s 12 Days event, where president Greg Brockman hailed it as a step function improvement on the hardest benchmarks.

The shift to reasoning at scale

Unlike GPT-4o, which focuses on raw speed and multimodal chat, reasoning models like o1 and o3 take a deliberate pause before answering. They run internal chain-of-thought checks, trading instant response times for rigorous accuracy in advanced coding, math, and scientific logic.

The o3-mini model shrinks that architecture down. It requires significantly less compute than full-sized models, which makes it cheap enough to serve at massive scale while delivering faster response times than heavier reasoning models.

Historically, OpenAI pushes older models down to the free tier only after newer flagships have been out for months. Giving free users access to a brand-new reasoning model right out of the gate is a notable change in strategy, even if OpenAI has yet to commit to a specific launch date.

Why it matters

We think this is one of OpenAI’s sharpest distribution moves in months, and it effectively resets the baseline for what a free AI tool should be capable of.

For everyday users and students, getting access to genuine reasoning without a 20-dollar monthly paywall is huge. Tasks that usually trip up free generative models, such as debugging tricky code or solving multi-step logic proofs, will suddenly become reliable. It makes advanced problem-solving accessible to millions who would never swipe a credit card for ChatGPT Plus.

For the broader industry, it turns advanced reasoning into a commodity. Rivals that are still charging premium prices for basic logic capabilities will find themselves squeezed from below. If the default free tier everywhere can reason through tough STEM problems, paying for incremental chatbot improvements becomes a much harder sell.

The trade-off, of course, will come down to compute bottlenecks. Even a lightweight reasoning model consumes far more compute per query than a standard autoregressive model when thinking through problems. OpenAI is betting that o3-mini’s smaller footprint will keep server infrastructure intact while creating a massive pipeline of users ready to upgrade when they hit usage caps.