GPT 5.4 Nano

GPT-5.4 Nano delivers flagship-class AI for classification, extraction, and high-volume API workloads at the lowest cost.

~3.29s
~$0
Playground
APIPricing

Chat

0 messages

Press Enter to send, Shift + Enter for new line • Max 5 files (10MB each)

GPT-5.4 Nano — Lightweight Text & Vision Language Model

What is GPT-5.4 Nano?

GPT-5.4 Nano is OpenAI's smallest and most cost-efficient model in the GPT-5.4 family, released on March 17, 2026. Built exclusively for API access, it is designed for developers and product teams running high-volume, latency-sensitive workloads where every token counts. With a 400,000-token context window, support for text and image inputs, and pricing starting at $0.20 per million input tokens, Nano delivers GPT-5.4-class intelligence at the lowest cost in its class — outperforming the previous generation's GPT-5 Mini at maximum reasoning effort.

Unlike consumer-facing models, GPT-5.4 Nano is purpose-built for infrastructure roles: classification agents, data extraction pipelines, ranking systems, and coding subagents. It is not available through ChatGPT; access is via the OpenAI API and compatible platforms like Segmind.

Key Features

  • •400K context window — process long documents, transcripts, and codebases in a single pass
  • •Image input support — analyze screenshots, charts, forms, and photos alongside text prompts
  • •Function calling & structured outputs — reliable JSON schema adherence for agentic pipelines
  • •Reasoning token support — enhanced accuracy on multi-step classification and logic tasks
  • •Distillation capability — generate synthetic training data from GPT-5.4 Nano outputs
  • •Batch API discount — 50% off pricing for asynchronous, non-time-critical workloads
  • •Tool integrations — web search, file search, code interpreter, MCP, and more via Responses API

Best Use Cases

GPT-5.4 Nano shines in scenarios that require speed, volume, and cost control over raw capability depth. Use it for text classification pipelines where you need to categorize thousands of documents per hour; data extraction tasks like pulling structured fields from unstructured invoices, contracts, or support tickets; and ranking and scoring at scale, such as relevance scoring or content moderation queues.

It is well-suited as a coding subagent handling simpler supporting functions within a larger multi-agent system — triaging issues, auto-labeling PRs, or summarizing test logs. For batch jobs with asynchronous workloads (nightly enrichment, bulk tagging, large-scale translation), the 50% Batch API discount makes Nano the most economical option available.

Prompt Tips and Output Quality

GPT-5.4 Nano responds best to clear, scoped instructions. For classification tasks, enumerate your categories explicitly in the prompt and ask for JSON output with structured outputs enabled. For extraction, describe the output schema precisely — field names, types, and examples.

For image tasks, describe what to look for in text alongside the image; Nano handles visual grounding for charts and documents well, but complex spatial reasoning should be routed to GPT-5.4 Mini instead. Keep prompts under 2,000 tokens where possible for maximum throughput efficiency.

Avoid open-ended creative or architectural tasks — Nano performs best on bounded, well-defined problems. When using function calling, keep your tool schemas lean and focused.

FAQs

Is GPT-5.4 Nano available in ChatGPT? No. GPT-5.4 Nano is API-only and does not appear in the ChatGPT consumer interface. Access it through the OpenAI API or Segmind.

Does GPT-5.4 Nano support image inputs? Yes. It accepts text and image inputs. Audio and video are not supported.

What is the context window for GPT-5.4 Nano? 400,000 tokens input, with a maximum of 128,000 tokens output.

How does GPT-5.4 Nano compare to GPT-5.4 Mini? Nano is faster and cheaper but lower capability. Mini achieves 72.1% on OSWorld-Verified vs Nano's 39.0%. Use Nano for simple, high-volume tasks and Mini for complex reasoning or multimodal tasks.

Is there a batch discount? Yes — the Batch API offers a 50% discount for asynchronous workloads processed within 24 hours.

What tools does GPT-5.4 Nano support? Via the Responses API: web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, and MCP integrations. Computer use is not supported.