What is the story about?
OpenAI has expanded the capabilities of its free ChatGPT tier with the rollout of its new GPT-5.4 Mini model, alongside the smaller GPT-5.4 Nano, just weeks after introducing its flagship GPT-5.4 system.
The move is aimed at making advanced AI tools faster and more accessible, particularly for users who prioritise speed and efficiency over raw power.
The GPT-5.4 Mini model is now available to free and Go users, marking a notable upgrade to the everyday ChatGPT experience. Designed as a lighter and more cost-effective alternative, it is positioned to handle a wide range of tasks while maintaining responsiveness.
GPT 5.4 Mini and Nano: How are they different
OpenAI says GPT-5.4 Mini delivers improved performance across coding, reasoning, multimodal understanding, and tool usage. The model is reported to run more than twice as fast as its predecessor, while still achieving performance levels close to the full GPT-5.4 on select benchmarks such as SWE-Bench Pro and OSWorld-Verified.
A key focus area is real-time interaction. The model is optimised for applications where latency matters, including coding assistants, systems that interpret screenshots, and tools that rely on quick, parallel processing.
It is also capable of analysing complex visual inputs, allowing it to understand dense user interfaces and reason over images with greater speed.
GPT-5.4 Nano, on the other hand, is positioned as the smallest and most economical option in the lineup. While more limited in scope, it is designed for lightweight tasks such as classification, data extraction, ranking, and supporting coding workflows.
Subagent role unlocks new workflows
One of the more notable additions is the introduction of “subagent” functionality. OpenAI is positioning GPT-5.4 Mini as a supporting model that can work alongside more powerful systems in multi-step workflows.
In developer environments such as Codex, GPT-5.4 can handle higher-level planning and decision-making, while delegating narrower, repetitive tasks to GPT-5.4 Mini. These can include scanning large codebases, reviewing documents, or executing parallel subtasks, helping improve overall efficiency.
The new model is also capable of handling complex computer-based tasks, including interpreting screenshots and interacting with software interfaces, narrowing the gap with the flagship model in visual performance.
GPT-5.4 Mini can be accessed through the ‘Thinking’ feature in ChatGPT’s interface, while developers can integrate both Mini and Nano models via OpenAI’s API. However, GPT-5.4 Nano will remain API-only for now.
