The Story

OpenAI has introduced GPT-5.4 mini and GPT-5.4 nano, expanding the GPT-5.4 family with smaller, faster, and more cost-efficient models built for coding, subagents, and high-volume production workloads.

According to the official announcement from OpenAI, GPT-5.4 mini brings many of GPT-5.4’s strengths into a model that runs more than twice as fast as GPT-5 mini, while GPT-5.4 nano is positioned as the smallest and cheapest option for simpler supporting tasks.

What the new models are built for

• Fast coding workflows: OpenAI says the models are optimized for targeted code edits, debugging loops, front-end generation, and codebase navigation.
• Subagents: GPT-5.4 mini is designed to work well in multi-model systems where a larger model handles planning while smaller agents execute narrower subtasks in parallel.
• Low-latency multimodal use: The models are also aimed at screenshot interpretation, tool use, and real-time image reasoning workloads.

Performance highlights

• GPT-5.4 mini: OpenAI reports that mini significantly improves over GPT-5 mini across coding, reasoning, multimodal understanding, and tool use, while approaching GPT-5.4 on several evaluations.
• Benchmark gains: On SWE-Bench Pro (Public), GPT-5.4 mini scored 54.4%, while GPT-5.4 nano scored 52.4%, both ahead of GPT-5 mini at 45.7%.
• Computer use: On OSWorld-Verified, GPT-5.4 mini reached 72.1%, close to GPT-5.4 at 75.0% and well above GPT-5 mini at 42.0%.

Availability and pricing

• GPT-5.4 mini: Available in the API, Codex, and ChatGPT. In the API it supports text and image inputs, tool use, function calling, web search, file search, computer use, and skills, with a 400k context window.
• Mini pricing: OpenAI lists GPT-5.4 mini at $0.75 per 1M input tokens and $4.50 per 1M output tokens.
• GPT-5.4 nano: Available only in the API at $0.20 per 1M input tokens and $1.25 per 1M output tokens.

Why it matters

As AI products move from one-off chat responses to always-on systems, smaller models are becoming more important for cost control, speed, and orchestration. OpenAI is clearly positioning GPT-5.4 mini and nano as the operational layer for fast coding assistants, support agents, and multimodal systems that need strong performance without flagship-model pricing.

The release also signals how frontier labs are increasingly building model families around specific workload tiers, where the winning product may depend as much on latency and economics as on raw intelligence.