OpenAI Unveils GPT-5.4 Mini and Nano: The Most Sophisticated Compact Models Ever

OpenAI Unveils GPT-5.4 Mini and Nano: The Most Sophisticated Compact Models Ever

2 Min Read

OpenAI has recently unveiled the launch of GPT-5.4 mini and nano, signifying a major leap forward in their series of AI models. These newly introduced models are described as the “most capable small models to date,” crafted to boost performance for users needing efficient processing for high-throughput tasks.

### GPT-5.4 Mini

The GPT-5.4 mini model brings notable enhancements compared to its predecessor, GPT-5 mini. It shines in multiple areas including programming, logical reasoning, multimodal comprehension, and tool utilization, all while functioning at more than double the speed of the earlier version. Significantly, it nears the performance benchmarks of the larger GPT-5.4 model in various assessments, such as SWE-Bench Pro and OSWorld-Verified, making it a powerful option for users seeking efficiency without sacrificing capability.

### GPT-5.4 Nano

Conversely, GPT-5.4 nano is crafted as the most compact and budget-friendly variant of the GPT-5.4 lineup. It is especially optimized for scenarios where both speed and cost efficiency are essential. This model signifies a remarkable advancement over GPT-5 nano and is ideal for tasks like classification, data extraction, ranking, and simpler programming assignments that require lower computational resources.

### Availability

As of today, GPT-5.4 mini is available to users of ChatGPT, Codex, and OpenAI’s API. ChatGPT users can locate it in the “Thinking” feature within the + menu for Free and Go users. Other users can utilize GPT-5.4 mini as a backup option for GPT-5.4 Thinking.

Meanwhile, GPT-5.4 nano is obtainable through the OpenAI API, facilitating developers to incorporate its functionalities into their applications.

### Recent Updates

Alongside the launch of these models, OpenAI has also rolled out GPT-5.4 Thinking, which features six pivotal enhancements geared towards improving user engagement. Earlier in March, the organization introduced GPT-5.3 Instant, aimed at mitigating occurrences of undesirable output, often termed “cringe.” Additionally, OpenAI released its Codex development application for Mac in February, broadening its range of tools available for developers.

For further detailed information about how these innovative models enhance programming, agentic workflows, and computer usage, users can consult OpenAI’s official releases and documentation.

You might also like