OpenAI has introduced two new compact models—GPT-5.4 mini and GPT-5.4 nano—expanding its GPT-5.4 lineup with a focus on speed and affordability.
The move reflects a shift toward practical performance. Instead of prioritising model size, these versions are designed for environments where faster responses and lower costs matter more, such as coding assistants, real-time apps, and multi-agent systems.
GPT-5.4 mini delivers a noticeable upgrade over GPT-5 mini. It performs better across coding, reasoning, multimodal tasks, and tool use, while running more than twice as fast. Despite being smaller, it comes close to the full GPT-5.4 model on key benchmarks. For example, on SWE-Bench Pro, it scores 54.4% accuracy, compared to 57.7% for GPT-5.4, and clearly ahead of GPT-5 mini’s 45.7%. It also shows strong results in tests like Toolathlon and GPQA Diamond, indicating that it can handle demanding tasks efficiently without the higher cost.
GPT-5.4 nano is the most lightweight option in the lineup. It is built for simpler, large-scale tasks such as classification, ranking, data extraction, and basic coding functions. While it doesn’t match the performance of the mini model, it offers much lower pricing, making it suitable for high-volume use cases.
The company is also highlighting a broader change in how AI systems are being built. Instead of relying on a single powerful model, developers are increasingly using multi-agent setups. In this approach, a larger model like GPT-5.4 handles planning and coordination, while smaller models such as GPT-5.4 mini take on specific tasks like document analysis, file processing, or code searches. This allows multiple tasks to run at the same time, improving efficiency and reducing costs.
OpenAI noted that this modular approach will become more effective as smaller models continue to improve in speed and capability.
GPT-5.4 mini is available through OpenAI’s API, Codex, and ChatGPT. It supports text and image inputs, tool use, and long-context reasoning with a context window of up to 400,000 tokens. Pricing is set at $0.75 per million input tokens and $4.50 per million output tokens.
GPT-5.4 nano is also accessible via the API at lower pricing levels, making it a practical option for deployments where keeping costs low is a priority.
Also Read: OpenAI Acquires Promptfoo to Strengthen Security for Its AI Agents
