Google: Gemini 3.6 Flash (batch)

Google: Gemini 3.6 Flash (batch) is a text model from Google, offered through api14 under the model id google/gemini-3.6-flash:batch. Google positions Gemini 3.6 Flash as a high-efficiency model aimed at coding, agentic workflows, and web and app development, and it is built to produce polished outputs with fewer unnecessary edits. The batch variant is the version you call through api14 when you want this model's behaviour on the batch endpoint rather than an interactive one.

The model accepts text, image, video, file, and audio as inputs, and it returns text. That means you can send a prompt that mixes written instructions with a screenshot, a video clip, a document, or an audio recording, and read back a single text response. The context window is 1,048,576 tokens, which is large enough to hold long documents, extended conversations, or sizeable codebases in one request. Output is capped at 65,536 tokens per response, so very long generations need to be planned around that ceiling.

Those capabilities point to a fairly wide set of practical uses. Coding assistance fits naturally: reading a repository or a long file, reasoning about it, and writing or revising code. Agentic workflows are another fit, since the model supports tool calling and structured outputs, which lets it decide when to invoke a function and return results in a shape your program can parse. Web and app development tasks, such as turning a design or a specification into working markup or component logic, follow from the same combination. Because inputs can include images, video, audio, and files, you can also build pipelines that summarise or extract information from media and documents without a separate transcription or vision step.

Calling it through api14 works the same way as any other OpenAI-compatible model: send a request to the chat completions endpoint with the model field set to google/gemini-3.6-flash:batch. The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, stop, structured_outputs, tool_choice, and tools. In practice that means you can control output length with max_tokens, steer or expose the model's reasoning with the reasoning parameters, request JSON or schema-conforming output with response_format and structured_outputs, make sampling reproducible with seed, define stop sequences, and give the model callable functions through tools and tool_choice.

One thing to keep in mind is the output limit: 65,536 tokens per response is generous but finite, so tasks that require a single very long generation may need to be split. The model outputs text only, so if you need generated images, audio, or video, this is not the right choice. It is also a Flash-tier model, which Google describes as high-efficiency rather than its largest option, so workloads where maximum reasoning depth matters more than efficiency may be better served by a larger model in the same family.