OpenAI Accelerates with GPT-5.4 Mini and Nano to Dominate Coding

OpenAI Unveils Its New Ultra-Fast AI Models

OpenAI has recently introduced two new artificial intelligence models, GPT-5.4 mini and nano, designed to provide speed and efficiency for ChatGPT users and developers using the OpenAI API. These models aim to meet the needs of specific tasks such as vibe coding, where larger AI models can be too costly.

A Response to Competition

The launch of these models is part of OpenAI's strategy to focus on coding, a sector where it competes with Anthropic. The latter recently made waves with its Claude Code model, capable of creating complete applications. OpenAI's new models, integrated into its coding software Codex, aim to directly compete with this offering.

Enhanced Performance

GPT-5.4 mini stands out with an execution speed more than twice as fast as its predecessor, GPT-5 mini. It excels in tasks such as coding, reasoning, and tool usage, and approaches the performance of the standard model GPT-5.4 on certain benchmarks.

OpenAI recommends GPT-5.4 mini for tasks like code editing and debugging. It can function as a sub-agent within Codex, allowing a larger model like GPT-5.4 to delegate certain tasks to this faster and more cost-effective model.

A Diverse Offering

GPT-5.4 nano, on the other hand, is even more compact and is recommended for basic tasks such as classification and data extraction. This model is exclusively available via the API.

Developers can access GPT-5.4 mini through the API, as well as via Codex and ChatGPT. Users of ChatGPT Free and Go will be able to use this model through the "Thinking" feature, while others will find it as an alternative when the rate limit for GPT-5.4 Thinking is reached. In contrast, GPT-5.4 nano remains accessible only via the API.