LLM & ChatActive

GPT-5 Nano AI Model

GPT-5 Nano is a GPT-5 model that OpenAI describes as a good fit for summarization and classification. It reads text and up to 10 images and supports function calling, web search and structured output.

Partner LLMLLM Tool CallLLM Response Format

Model ID

gpt-5-nano

Provider

openai

Updated

Oct 23, 2026, 21:13

Support Wiro on G2

wiro playground—openai/gpt-5-nano

Updated Oct 23, 2026, 21:13

Overview

GPT-5 Nano belongs to OpenAI's GPT-5 family, and OpenAI says it is well suited to summarization and classification. It takes text and image input, returns text, and can spend reasoning tokens before it answers. OpenAI lists a 400,000-token context window (up to 272,000 input tokens and 128,000 output tokens) and a knowledge cutoff of May 31, 2024. Its dated snapshot is gpt-5-nano-2025-08-07.

This page exposes the GPT-5 controls for reasoning effort and verbosity, plus function tools, structured response formats and optional web search.

What you can build

Ticket, email or review triage that returns labels in a fixed JSON schema
Summaries of long threads, reports and call transcripts
Answers drawn from screenshots, receipts or product photos
Agents that call your own functions, several in one turn when needed
Replies about recent events, with web search switched on

Inputs

Text goes in as a single prompt or as a messages array holding the whole conversation. Use one of the two, never both. Up to 10 images in JPG, PNG, WebP or non-animated GIF can be attached alongside. System instructions set the model's behavior for the run. For stored history, give the run a user ID and a session ID; the page can generate random values for both.

Reasoning effort is minimal, low, medium (default) or high. Verbosity is low, medium (default) or high and controls how long and detailed the answer gets. Max output tokens accepts up to 65,536. Web search is off by default, and when it is on, the number of searches depends on how complex the prompt is.

For tool use, add function definitions, set tool choice to auto, none, required or a specific tool, and allow or block parallel tool calls (allowed by default). When the model asks for a tool, send the results back in the tool outputs field together with the previous run's task token. The response format field requests structured output.

Outputs

Text answers by default. With tools defined, the reply can instead be one or more function calls carrying JSON arguments, and with a response format set, the answer follows that structure.

Recommended settings

Labeling and extraction: minimal or low reasoning, low verbosity and a response format, so replies stay short and machine-readable.
Multi-step questions, or reading fine detail in images: medium or high reasoning.
Web search does not work with minimal reasoning, so choose low or above when search is on.

Limitations

The upload field takes images only, so PDFs and other documents cannot be attached here, even though OpenAI lists file uploads for this model.
Knowledge stops at May 31, 2024. Later events need web search.
Max output tokens takes values up to 65,536, about half the 128,000 OpenAI lists. Left empty, as it is by default, the field sets no limit, so OpenAI's own 128,000-token ceiling applies.

Safety & compliance

If an image shows a recognizable person, get that person's consent before you send it.

API quick start

Run gpt-5-nano with a single API call.

POST https://api.wiro.ai/v1/Run/openai/gpt-5-nano

{
  "prompt": "Tell me about Türkiye",
  "inputImage": "https://your-cdn.com/input.png",
  "user_id": "...",
  "session_id": "..."
}

curl

curl -X POST "https://api.wiro.ai/v1/Run/openai/gpt-5-nano" \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_WIRO_API_KEY" \
  --data-binary @- <<'JSON'
{
  "prompt": "Tell me about Türkiye",
  "inputImage": "https://your-cdn.com/input.png",
  "user_id": "...",
  "session_id": "..."
}
JSON