OpenAI released GPT-5.4 mini and GPT-5.4 nano on March 17, bringing the capabilities of its flagship GPT-5.4 model to significantly smaller, faster, and cheaper packages. Mini runs more than 2x faster than GPT-5.4 and approaches its performance on coding and reasoning benchmarks. Nano is the company's smallest and cheapest model to date.

What Happened

GPT-5.4 mini is now available to ChatGPT Free and Go users through the "Thinking" menu option. On the API side, it supports text and image inputs, tool use, and a 400,000-token context window at $0.75 per million input tokens and $4.50 per million output tokens. That makes it roughly 10x cheaper than GPT-5.4 while retaining most of its capabilities.

GPT-5.4 nano is API-only, priced at just $0.20 per million input tokens ($1.25 output). OpenAI positions it for classification, data extraction, ranking, and lightweight coding tasks that run as subagents within larger systems.

Why It Matters

Cheaper, faster models directly affect what creators can build. At $0.20 per million tokens, nano makes it practical to run AI-powered workflows that would have been cost-prohibitive even six months ago. Think real-time content tagging, automated metadata generation, or AI assistants embedded in creative tools that need to respond instantly without burning through API budgets.

The subagent architecture is especially relevant. In OpenAI's own Codex system, GPT-5.4 handles planning and coordination while mini subagents tackle narrower tasks in parallel. This pattern is already spreading to creative tools where a main model directs specialized agents for image generation, copywriting, and layout.

Key Details

  • GPT-5.4 mini: $0.75/1M input, $4.50/1M output, 400k context, 2x faster than GPT-5.4
  • GPT-5.4 nano: $0.20/1M input, $1.25/1M output, API-only
  • Mini matches GPT-5.4 on SWE-Bench Pro and OSWorld-Verified benchmarks
  • Mini available in ChatGPT Free tier (select "Thinking" from plus menu)
  • Both models support tool use and image inputs

What to Do Next

If you're building AI-powered creative tools or workflows, test mini as a drop-in replacement for GPT-5.4 in your API calls. The 400k context window means you can process entire documents, codebases, or long creative briefs in a single request. For high-volume tasks like content classification or metadata extraction, nano at $0.20/1M tokens makes batch processing nearly free. The GPT-5.4 computer use mode launched earlier this month pairs well with these smaller models for building multi-agent creative workflows.