The stable model ID in the Gemini API is gemini-4-neon. It accepts text, images, video, audio, and PDFs, and returns text. Input and output share a budget of 1,048,576 tokens, and the default output limit for synchronous calls is 65,536 tokens.
Suitable tasks. Use the Google Gemini API to turn text or multimedia into text or structured output.
Unsuitable tasks. Direct image or audio generation, or real-time interaction through the Live API.
Supported model version. The stable Gemini 4 Neon release, gemini-4-neon.
Supported client, agent, or API. Google Gemini API. The example below uses the Interactions API's REST interface.
Recommended thinking levels and parameters. The model specifications confirm support for Thinking. Available levels and default values are unspecified. The minimal request omits the thinking configuration.
This request follows the REST request structure in Google's general quickstart, using the model ID and input for this example. Gemini API quickstart REST example
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-H "Api-Revision: 2026-05-20" \
-d '{
"model": "gemini-4-neon",
"input": "Explain the difference between CSV and JSON in one sentence."
}'Set GEMINI_API_KEY locally. Use the text request above to confirm that your account can call gemini-4-neon and receive text in response.
Add images, video, audio, or PDFs as your use case requires, and check the text output for each input type. If you need structured output, validate the fields against your application's schema.
Leave room in the context window for long outputs. Input tokens + output tokens ≤ 1,048,576. For outputs above the default synchronous limit, follow the official guide to configure long decode. Gemini API long decode guide
| Specification | Value listed on Google's model page |
|---|---|
| Stable model ID | gemini-4-neon |
| Input / output | Text, images, video, audio, PDFs / text |
| Combined input and output limit | 1,048,576 tokens |
| Default synchronous output limit | 65,536 tokens |
| Long decode output limit | Up to 1,048,576 tokens, subject to the context used by the input |
| Tools and output capabilities | Agentic video understanding, code execution, computer use, file search, function calling, grounding with Google Maps and Google Search, structured output, Thinking, and URL context |
| Caching | Context caching is supported. Explicit caching is unavailable. |
| Unsupported capabilities | Audio generation, image generation, Live API |
| Inference service options | Batch API, Flex inference, Priority inference |
The model specifications omit file size limits and supported formats for each input type, as well as limits on combining tools.
For pricing, regions, account permissions, and quotas, refer to the relevant official pages and your project configuration.
On 2026-11-19, I used the international edition of Tabbit to read Google's official model specifications and REST quickstart. I assembled the example from the official documentation without calling the Gemini API.
Gemini 4 Neon