The baseline just moved.
OpenAI confirmed that GPT-5.4 Mini is now available to Free and Go users through the “Thinking” feature in ChatGPT’s + menu. According to OpenAI’s official announcement, the rollout happened March 18. Free users don’t pick a model name. They pick a reasoning level, and GPT-5.4 Mini is what powers that experience.
For paid users, the mechanics are different. Plus, Pro, Business, Enterprise, and Edu accounts see GPT-5.4 Mini as a rate-limit fallback, it activates automatically when GPT-5.4 Thinking capacity is exhausted. The same announcement confirmed that ChatGPT’s model picker was updated for paid tiers to organize options around reasoning levels rather than individual model names. Enterprise customers may also configure GPT-5.4 Mini as a default via auto-routing, according to OpenAI, though this detail wasn’t present in the source excerpts available for this brief.
OpenAI describes GPT-5.4 Mini as offering enhanced reasoning capabilities suitable for general queries. No independent benchmark data has been published. No arXiv paper is available. Treat performance characterizations as OpenAI’s positioning until third-party evaluation exists.
One item not covered in this brief: 9to5Google and OpenAI’s own announcement title both reference GPT-5.4 Nano, released simultaneously. That model isn’t covered here, see the coverage gap note in the production log. It’s worth knowing it exists before building anything that assumes Mini is the only new model in this family.
The competitive implication is direct. When reasoning-capable AI is free, products that charge for “smarter AI” need a sharper value proposition than model tier alone.