OpenAI has launched GPT-5.4 mini and nano, new AI models optimized for fast performance in coding and multimodal applications.
OpenAI has unveiled its latest AI models, GPT-5.4 mini and GPT-5.4 nano, designed for applications that require rapid response times.
Introducing GPT-5.4 mini and nano, our most advanced compact models to date. The mini version is over twice as fast as its predecessor, GPT-5 mini, and is tailored for coding, user interfaces, multimodal comprehension, and subagent tasks.
— OpenAI Developers (@OpenAIDevs) March 17, 2026
The new mini model excels in coding, logical reasoning, and integration with external tools, boasting more than double the data processing speed when compared to the GPT-5 mini. It has shown impressive results in tests such as SWE-Bench Pro and OSWorld-Verified, nearing the performance levels of GPT-5.4.
GPT-5.4 mini is crafted for scenarios where quick responses enhance product quality.
Effective applications for these new models include:
- AI assistants that require immediate responses;
- Subagents executing tasks rapidly;
- Systems managing interfaces that interpret screenshots;
- Multimodal applications for real-time image analysis.
In many cases, the ideal model prioritizes speed over sheer power, delivering prompt answers and efficiently managing tools while handling complex tasks.
GPT-5.4 nano, the most compact model, is recommended for basic tasks like data classification, extraction, and ranking.
Programming Enhancement
The mini version is particularly suitable for development tasks requiring quick iterations, demonstrating superiority in pinpoint editing, navigating code, frontend generation, and debugging.
This makes it a prime choice for projects that demand both speed and cost-effectiveness.
Benchmarks indicate that GPT-5.4 mini consistently outperforms GPT-5 mini, delivering faster response times.
Efficiency in Computer Use
In multimodal tasks, GPT-5.4 mini effectively manages user interfaces, interpreting complex screenshots for rapid computer actions.
Access and Cost Structure
The GPT-5.4 mini model is accessible through API, Codex, and ChatGPT, with a pricing structure of $0.75 per million input tokens and $4.50 per million output tokens. In contrast, the nano version is exclusively available via API at lower rates of $0.20 for input tokens and $1.25 for output tokens. OpenAI has also integrated the music recognition service Shazam into ChatGPT, allowing users to identify songs seamlessly.
