Google: Gemini 3.1 Pro Preview Custom Tools

Google: Gemini 3.1 Pro Preview Custom Tools is a text model from Google, offered as a preview variant of Gemini 3.1 Pro. Its distinguishing trait is a change to tool selection behavior: the model is tuned to avoid overusing a general bash tool when more efficient third party options are available. The provider released it on 2026-02-25, and it sits in the text category on api14.

The model accepts text, audio, image, video and file inputs, and it returns text only. Its context window is 1,048,576 tokens, with a maximum output of 65,536 tokens. Reasoning behavior is exposed through the reasoning, reasoning_effort and include_reasoning parameters, so you can control how much internal deliberation the model applies and whether that reasoning is surfaced in the response.

Those capabilities point to a few practical patterns. The very large context window suits work over long documents, transcripts or multi file inputs, and the mix of input modalities means you can pass audio, images, video or files alongside text instructions without a separate preprocessing step. Because tool selection is the focus of this variant, it fits agentic workflows where the model must choose among several tools, particularly when a general bash tool tends to dominate in other configurations. Structured outputs and reasoning support also make it reasonable for extraction and analysis tasks where you want a predictable response shape.

You call it through api14's OpenAI-compatible API using the model id google/gemini-3.1-pro-preview-customtools. Requests support include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, structured_outputs, temperature, tool_choice, tools and top_p. If you want a tool bearing agent, pass tools and tool_choice; if you want deterministic-ish sampling, use seed and temperature. For structured data, use response_format or structured_outputs. Long inputs are worth checking against the 1,048,576 token context and the 65,536 token output cap.

Keep in mind that this is a preview build and a specialized variant, so it is worth testing against your own toolset rather than assuming the tool selection tuning matches your setup. Output is text only, so anything needing generated images, audio or video would require a different model. If your work does not involve tool selection, a general purpose Gemini release may be a simpler fit, since the custom tools behavior is the main reason to pick this one.