OpenAI Unveils GPT-5.4 Mini and Nano: Approaching Flagship Performance, Priced Up to 4x the Predecessor

According to 1M AI News, OpenAI has released GPT-5.4 mini and GPT-5.4 nano, targeting Agent, Programming, and Multi-Modal Workflow scenarios. The API pricing for GPT-5.4 mini is $0.75 per input and $4.50 per output (per million tokens), which is 2.25 to 3 times that of the previous GPT-5 mini; nano is priced at $0.20 per input and $1.25 per output, up to 4 times that of the previous generation.

In terms of performance, GPT-5.4 mini achieves 54.4% on the SWE-Bench Pro programming benchmark, close to GPT-5.4's 57.7%, significantly outperforming GPT-5 mini's 45.7%. Both models support a context window of 400,000 tokens. mini supports functionalities such as programmatic control, tool invocation, and web search, with a runtime speed of over twice that of the previous generation.

ChatGPT Free and Go users will have access to the Inference feature (Thinking) for the first time, powered by GPT-5.4 mini. The mini is simultaneously integrated with GitHub Copilot and Codex, consuming only 30% of the GPT-5.4 quota in Codex. The nano is only available via the API, suitable for lightweight tasks such as classification and data extraction.