GPT-5.4 Mini: Fast AI, Half the Cost

OpenAI Launches GPT-5.4 Mini and Nano: The Most Capable Small Models Yet

OpenAI has just released GPT-5.4 mini and nano, pushing the boundaries of efficient AI with models that deliver superior performance in speed and capability. These new small models are designed for high-volume tasks, offering significant upgrades in coding, reasoning, and multimodal understanding while running faster and costing less.

What Are GPT-5.4 Mini and Nano?

GPT-5.4 mini stands out as OpenAI's strongest mini model to date. It excels in coding, reasoning, multimodal understanding, and tool use, surpassing GPT-5 mini while operating more than twice as fast. This model approaches the performance of the full GPT-5.4 on key benchmarks like SWE-Bench Pro and OSWorld-Verified.

GPT-5.4 nano takes efficiency further as the smallest and cheapest version in the GPT-5.4 family. Ideal for tasks prioritizing speed and cost, such as classification, data extraction, ranking, and simpler coding subagents, it represents a major upgrade over GPT-5 nano.

Key Improvements and Capabilities

Both models build on the GPT-5.4 foundation, which integrates advances in reasoning, coding, and agentic workflows. GPT-5.4 mini supports a 400,000 token context window, tools like functions, web search, file search, and computer use, with input pricing at $0.75 per million tokens and output at $4.50 per million tokens.

• Enhanced speed for high-volume workloads.
• Strong performance in computer use and browser interactions.
• Support for text and image inputs, multilingual capabilities, and vision.
• Max output of 128,000 tokens for mini.

These models are optimized for latency and cost, making them perfect alternatives to larger variants for everyday and professional applications.

Availability and Access

ChatGPT users can access GPT-5.4 mini immediately. Free and Go users find it via the "Thinking" feature in the + menu, while others use it as a rate limit fallback for GPT-5.4 Thinking. GPT-5.4 nano is available through the OpenAI API, with tiered limits for requests per minute and tokens per minute.

Developers can integrate these via the API, Assistants, Batch, and more, starting with models suited for complex reasoning or lighter workloads.

Context Within OpenAI's GPT-5 Series

The release follows GPT-5.4, launched earlier in March 2026 as OpenAI's flagship for professional work. It features a 1 million token context window, 33% fewer factual errors, and native computer use capabilities like controlling desktops via screenshots and inputs. GPT-5.4 also shines in benchmarks, achieving top scores in OSWorld-Verified, WebArena-Verified, and knowledge work tasks.

Recent additions include GPT-5.4 Thinking for advanced reasoning and GPT-5.3 Instant for quick daily tasks, expanding options across free, Plus, Pro, and Enterprise tiers.

Conclusion

GPT-5.4 mini and nano mark a leap in accessible, high-performance AI, enabling developers and users to tackle demanding tasks efficiently without compromising quality.

These models signal OpenAI's commitment to balancing power, speed, and affordability, opening new possibilities for applications in coding, agents, and beyond. Stay tuned for how they transform workflows.