OpenAI Launches GPT-5.4 Mini and Nano to Provide Answers 2X Faster
OpenAI has officially launched GPT-5.4 Mini and GPT-5.4 Nano, introducing its most advanced small models designed for high-volume, latency-sensitive tasks.
OpenAI has officially launched GPT-5.4 Mini and GPT-5.4 Nano, introducing its most advanced small models designed for high-volume, latency-sensitive tasks.
The GPT-5.4 Mini offers a significant performance enhancement over previous iterations in areas such as reasoning, coding, tool use, and multimodal understanding, with the capability of operating more than twice as fast.
These models are tailored for applications where speed is crucial to user experience, including responsive coding assistants, real-time multimodal applications, and systems for capturing and interpreting screenshots.
The release demonstrates that larger models are not always optimal; systems that respond promptly and reliably are often more suitable for complex professional workflows.
Technical Specifications of GPT-5.4 Mini and Nano
Both models exhibit high efficiency in coding environments requiring rapid iteration, such as codebase navigation, debugging loops, generating front-end code, and executing targeted edits.
Benchmark results indicate that GPT-5.4 Mini approaches the accuracy of the flagship GPT-5.4 model on evaluations such as SWE-Bench Pro, providing one of the best performance-per-latency tradeoffs for developers.
OpenAI has officially launched GPT-5.4 Mini and GPT-5.4 Nano, introducing its most advanced small models designed for high-volume, latency-sensitive tasks.
A key technical feature is their integration into subagent architectures. On platforms like Codex, developers can utilize a larger model like GPT-5.4 for complex planning and judgment, while delegating specific tasks to GPT-5.4 Mini subagents.
These smaller agents can handle supporting documents, search codebases, and review large files concurrently, enhancing system efficiency and scalability.
GPT-5.4 Mini delivers considerable improvements in multimodal tasks, especially in computer-use scenarios. It can quickly analyze dense user interface screenshots to perform actions with high precision.
On the OSWorld-Verified benchmark, GPT-5.4 Mini achieved an accuracy of 72.1 percent, nearly matching the 75.0 percent score of the larger GPT-5.4 and significantly surpassing the 42.0 percent of the older GPT-5 Mini.
For simpler support tasks, GPT-5.4 Nano offers a cost-effective solution. It is recommended for data extraction, classification, ranking, and lightweight coding tasks where speed and cost efficiency are critical.
GPT-5.4 Mini is available via the OpenAI API, Codex, and ChatGPT. Within the API, it supports a 400k context window, allowing for text and image inputs, function calling, web search, and computer use. Pricing is set at $0.75 per one million input tokens and $4.50 per one million output tokens.
In Codex, developers can manage routine coding tasks using GPT-5.4 Mini at approximately one-third the cost, utilizing only 30 percent of the standard GPT-5.4 quota.
ChatGPT Free and Go users can access the model via the Thinking feature, which also serves as a rate limit fallback for other tiers.
The Nano variant is available exclusively through the API, priced at $0.20 per one million input tokens and $1.25 per one million output tokens.
Based on reporting by Cyber Security News.