OpenAI releases GPT-4o mini
Priced at 15 cents per million input tokens, more than 60% cheaper than GPT-3.5 Turbo, and became the default model for free ChatGPT users.
- Models & capabilities
- Minor
OpenAI released GPT-4o mini, a smaller and cheaper variant of its GPT-4o model, priced at roughly 15 cents per million input tokens and 60 cents per million output tokens — more than 60% below the cost of GPT-3.5 Turbo, the model it was built to replace. The company reported an MMLU score around 82%, ahead of rival small models it compared itself against, including Google’s Gemini Flash and Anthropic’s Claude Haiku.
GPT-4o mini inherited GPT-4o’s multimodal input handling and became the default model for ChatGPT’s free tier, while replacing GPT-3.5 Turbo as an option for Plus and Team subscribers and in the API. OpenAI positioned it as the model most developers building cost-sensitive applications should reach for by default, rather than the smallest available option — a shift from treating the cheapest tier as a stripped-down fallback to treating it as a genuinely capable model priced for high-volume use.
The release intensified competition at the lower end of the market, where Google, Anthropic and open-weight labs were all racing to cut the price of “good enough” intelligence, with per-token cost increasingly treated as a headline competitive metric alongside benchmark scores.