OpenAI has introduced two new models in the GPT-5.4 family: GPT-5.4 mini and GPT-5.4 nano. GPT-5.4 Mini is incorporated into the API, Codex, and ChatGPT platforms, while GPT-5.4 Nano is only available via the API. Both methods are intended to improve fast coding operations while minimising delay in task execution. GPT-5.4 mini outperforms the previous GPT-5 mini on a variety of tasks and can handle subagent tasks.
The OpenAI blog post said, “Today we’re releasing GPT‑5.4 mini and nano, our most capable small models yet. They bring many of the strengths of GPT‑5.4 to faster, more efficient models designed for high-volume workloads.”
In This Article
Pricing and Availability
GPT 5.4 Mini is now available on several platforms, including API, Codex, and ChatGPT. The API supports text and image input, tool use, function calling, web and file search, and has a huge context window of 400k. The price is set at $0.75 per million input tokens and $4.50 per million output tokens.
GPT-5.4 Mini is accessible via the Codex platform, including the app, CLI, IDE extension, and online. It uses just 30% of the GPT-5.4 quota, making it a cost-effective solution for basic programming. ChatGPT users may access this version via the “Thinking” function for Free and Go user tiers, while it acts as a backup alternative for those having access to GPT 5.4 Thinking. Additionally, GPT-5.4 nano is only accessible via the API at a cheaper cost of $0.20 per million input tokens and $1.25 per million output tokens.
Features
According to the business, GPT-5.4 mini outperforms GPT-5 mini in areas like coding, reasoning, multimodal understanding, and tool utilisation, with more than twice the performance speed. Its assessment findings, which include benchmarks such as SWE-Bench Pro and OSWorld-Verified, show that it approaches the performance standards established by the larger GPT-5.4 model. The GPT-5.4 nano model is recognised as the most compact and cost-efficient choice, particularly intended for applications requiring speed and cost-effectiveness, and it represents a major improvement over the GPT-5 nano. It is especially well-suited to applications like classification, data extraction, ranking, and basic coding tasks, making it perfect for situations that need low-latency improvements. This capability makes it suitable for responsive coding assistants, quick-support subagents, screenshot-interpreting systems, and real-time multimodal environments.
Also Read: Oppo A6s 5G launched in India at this pricing
Both models are especially intended to improve coding operations by allowing for fast iterations. They are especially useful in targeted edits, navigating codebases, generating front-end code, and executing debugging loops with minimal latency. These models provide a cost-effective option that considerably accelerates coding processes. Benchmark findings show that GPT-5.4 mini outperforms GPT-5 mini at similar latency levels while attaining pass rates comparable to GPT-5.4, resulting in remarkable performance relative to latency for coding applications.
Also Read: Oppo K14 5G with a 7,000mAh battery launched in India
GPT-5.4 Mini is intended for systems that use models of varying sizes, allowing bigger models, such as GPT-5.4, to handle planning, coordination, and final decisions. Smaller subagents, such as the GPT-5.4 mini, carry out particular tasks simultaneously, such as codebase searches and file reviews. The efficiency of these smaller models grows with their speed and capability, enabling developers to design systems in which bigger models lead tasks and smaller models perform them quickly. 5.4 mini distinguishes itself as the most advanced mini model designed specifically for this process.
GPT-5.4 Mini excels at multimodal activities, notably computer usage, where it can accurately understand screenshots of complicated user interfaces. It roughly equals GPT-5.4 on the OSWorld-Verified benchmark and outperforms GPT-5 mini.
