OpenAI's GPT-5.4 Mini Is Now Free — and It's Surprisingly Good

OpenAI released two new AI models this week — GPT-5.4 mini and GPT-5.4 nano — and the headline is that free ChatGPT users are getting an immediate upgrade without paying a cent.

What’s New

GPT-5.4 mini is OpenAI’s latest “small but capable” model. It runs more than twice as fast as the previous GPT-5.0 mini, and it’s meaningfully smarter — particularly at coding, reasoning, and understanding images and screenshots. On the GPQA Diamond benchmark (a test of expert-level reasoning), it scored 88%, which is remarkably close to the full flagship model’s 93%.

GPT-5.4 nano is the other half of the launch. It’s even smaller and cheaper, but it’s not something you’ll use in ChatGPT directly. It’s designed for developers building AI systems where speed and cost matter more than raw intelligence — things like sorting emails, extracting data from forms, or running background classification tasks.

Who Gets What

If you’re a free ChatGPT user, you now have access to GPT-5.4 mini. You can select it from the model menu under “Thinking.” It’s a real upgrade over what free users had before.

If you’re a developer using the API, both models are available today. Mini costs $0.75 per million input tokens; nano comes in at just $0.20 — making it one of the cheapest capable models OpenAI has ever offered.

GPT-5.4 nano is also designed specifically for what OpenAI calls “subagents” — the smaller AI components that do supporting work inside larger automated pipelines.

What This Means for You

If you use ChatGPT on the free tier and have felt limited by slower, less capable responses, this is a genuine step forward — no subscription required. For developers, nano opens up high-volume use cases that were previously too expensive to run at scale.

OpenAI is clearly pushing smaller, faster models to stay competitive as rivals like Anthropic and Google have made similar efficiency gains. The gap between “flagship” and “free” is shrinking fast.

You can try GPT-5.4 mini now at chat.openai.com — just look for it in the model selector.