OpenAI has released GPT‑5.4 mini and GPT‑5.4 nano, the latest and most capable small models. These versions bring much of GPT‑5.4’s capabilities to faster, more efficient models designed for high-volume and latency-sensitive workloads.
GPT‑5.4 Mini and Nano
- GPT‑5.4 mini improves on GPT‑5 mini across coding, reasoning, multimodal understanding, and tool use, running over 2× faster. It approaches GPT‑5.4 performance on evaluations such as SWE-Bench Pro and OSWorld-Verified.
- GPT‑5.4 nano is the smallest and least expensive version of GPT‑5.4, optimized for tasks where speed and cost are critical. It is a significant upgrade over GPT‑5 nano and is suitable for classification, data extraction, ranking, and coding subagents handling simpler supporting tasks.
These models are built for workloads where latency directly affects performance, including coding assistants requiring quick responses, subagents completing supporting tasks in parallel, computer-using systems interpreting screenshots, and multimodal applications reasoning over images in real time.
Key Features
Coding
- Supports fast iteration workflows with targeted edits, codebase navigation, front-end generation, and debugging loops.
- GPT‑5.4 mini outperforms GPT‑5 mini at similar latencies and nears GPT‑5.4 pass rates, delivering an efficient performance-to-speed ratio.
Subagents
- Works well in multi-model systems where larger models handle planning and judgment, while GPT‑5.4 mini executes narrower subtasks, such as searching codebases, reviewing large files, or processing supporting documents.
- Enables scalable systems by delegating simpler or parallel tasks to smaller, faster models.
Computer Use & Multimodal Tasks
- Rapidly interprets dense user interface screenshots to perform tasks efficiently.
- On OSWorld-Verified, GPT‑5.4 mini approaches GPT‑5.4 performance while substantially exceeding GPT‑5 mini.
Pricing and Availability
| Model | Platform | Features | Pricing |
|---|---|---|---|
| GPT‑5.4 mini | API, Codex, ChatGPT | Text and image inputs, tool use, function calling, web/file search, computer use, skills | $0.75 per 1M input tokens, $4.50 per 1M output tokens |
| GPT‑5.4 nano | API only | Lightweight tasks, fast execution | $0.20 per 1M input tokens, $1.25 per 1M output tokens |
Additional Notes:
- In Codex, GPT‑5.4 mini uses 30% of GPT‑5.4 quota, making simpler tasks cheaper and enabling subagent delegation.
- In ChatGPT, Free and Go users can access GPT‑5.4 mini via the “Thinking” menu. For other users, it functions as a rate-limit fallback for GPT‑5.4 Thinking.