OpenAI Launches GPT-5.4 Mini and Nano AI Models

OpenAI launched two compact artificial intelligence models, GPT-5.4 Mini and GPT-5.4 Nano, on March 18, 2026. These variants are designed for high-volume, latency-sensitive workloads such as coding assistants, real-time image analysis, and screenshot interpretation, aiming to make agentic workflows more commercially viable by reducing inference costs.

GPT-5.4 Mini serves as a fast, cost-efficient alternative that runs more than twice as fast as its predecessor and approaches the performance of the flagship GPT-5.4. It is available to ChatGPT Free and Go users via a Thinking option, as well as through the OpenAI API and Codex. OpenAI is promoting a subagent architecture where the flagship GPT-5.4 manages overall planning and delegates parallel tasks—such as codebase searching or document reviewing—to the Mini model.

GPT-5.4 Nano is the smallest and most affordable version, optimized for simple operations including classification, data extraction, and ranking. Unlike the Mini model, Nano is available exclusively through the OpenAI API. These releases target a reduction in the trade-off between capability and affordability for developers implementing customer support bots and edge-based systems.

Early adoption includes Notion Labs, Inc., which integrated GPT-5.4 Mini to handle focused tasks and complex formatting. Notion's AI engineering lead, Abhisek Modi, reported that the Mini model matches or outperforms more expensive models for formatting tasks.