gemini-3-flash-preview

Name: gemini-3-flash-preview
Brand: Google

GoogleChatReasoningVision

Get your API key

A multimodal reasoning model balancing interactive efficiency and long-context understanding

Gemini 3 Flash Preview is Google's multimodal thinking model, designed for applications requiring interactive efficiency, long-material understanding, and coding collaboration. It can analyze text and visual information together, delivering results as text or structured output. It also natively supports video, audio, and PDF understanding, making it suitable for organizing complex materials into readable, actionable answers.

Try this model View pricing

GoogleModel brand

ChatModel type

Reasoning, visual understandingTask capabilities

Specifications and API features

Clarify capacity, inputs and outputs, and invocation methods before selecting a model.

Version positioning: Gemini 3 Flash Preview; invocation ID: gemini-3-flash-preview
Native input limit: 1,048,576 Token
Native output limit: 65,536 Token
Native inputs and outputs: Text, images, video, audio, and PDF input; text output
Reasoning and orchestration: Thinking, structured output, function calling
Text-and-image invocation: /gemini/chat/completions; messages supports text and image_url content blocks
Conversation invocation: /aichat2/conversations; supports conversation ID, image blocks, and file link blocks

Native capacity and modality describe the model's capability boundaries; this platform provides input organization, conversation management, and tool operations separately according to the entry point used.

Core Capabilities

Learn what gemini-3-flash-preview can bring to your work.

Understand visual and textual clues together

It not only describes images, but can also analyze them in combination with questions, background text, and multiple images. It is suitable for reading interface screenshots, understanding chart meanings, comparing design options, and organizing observations into descriptions or fields. When providing input, clearly specify the areas of focus and evaluation criteria; this makes it easier to get useful results than simply asking “what is in the image.”

Long materials and reasoning capabilities work together

A large native input capacity is suitable for accommodating long documents, code snippets, and continuous conversations, while Thinking is used to analyze constraints, sort out relationships, and formulate answers. You can ask it to list key evidence first, then provide revision suggestions or a summary; long context helps keep materials coherent, but does not mean that every detail can be detected without error.

Move from natural language to structured delivery

Structured output and function calling are supported, enabling materials to be converted into business fields, check results, or tool parameters. Applications can describe the target format through response_format and define executable actions through function definitions. The model is responsible for understanding and proposing calls, while actual execution, permission checks, and result write-back are handled by the application workflow.

Applicable Scenarios

Start with specific tasks to find where the model can be effective.

Screenshot-driven product reviews

Submit product page screenshots, requirement descriptions, and acceptance criteria, and let the model identify layout issues, copy ambiguities, and process gaps, delivering a revision checklist organized by page area. Placing images and text in the same message keeps suggestions close to the visible interface, making this suitable for prototype discussions and test issue classification rather than automatically operating a browser.

Code review and implementation discussions

Provide relevant code, error logs, and expected behavior, and let the model explain possible problem paths, propose modifications, and generate testing ideas. Its ability to handle long materials helps discuss multiple related fragments at once, and the deliverables can include patch drafts and validation steps; running code, installing dependencies, and submitting changes should still be done in a controlled environment.

File Q&A and ongoing analysis

Submit file links and questions through the conversation entry point, and use the file-reading workflow to organize report highlights, extract clauses, or compare materials. After receiving a conversation ID, you can continue asking follow-up questions to gradually create summaries, comparison tables, and to-do lists. File-link input is a workflow and should not be confused with the model natively receiving PDFs directly.

How to choose this model

Choose based on task complexity, input materials, and expected results.

Choose Flash for everyday complex tasks, evaluate Pro for deep challenges

When a task involves image and text understanding, code assistance, and multi-turn interaction, and processing efficiency is also important, Gemini 3 Flash Preview is worth evaluating. If the core task is high-difficulty reasoning or complex solution exploration, compare it with Gemini 3.1 Pro Preview using the same set of materials. The choice should depend on actual answer quality and workflow fit, rather than assuming version names correspond to fixed performance differences.

Choose image analysis and generation separately

This model excels at understanding images and producing text; it is not an image generation model. Choose it when you need to review screenshots, explain charts, or extract visual information; when you need to deliver new images or image editing results, choose the appropriate image model. gemini-3-flash-preview is also not equivalent to other Flash versions, so when migrating existing applications, retain representative tasks for regression testing.

Get started

From a small-scale task to production integration.

Prepare tasks and materials

Define the goal, required inputs, and output requirements, using real business examples as a starting point.

Try it in the API testing area

Open the trial page, confirm the parameters supported by this entry point, then submit a small-scale task to review the results.

Integrate according to the API documentation

Keep the complete model ID, use the request format specified in the documentation, and confirm billing rules on the Pricing page.

Usage limits

Before formal use, understand output quality and capability scope.

Audio input understanding does not equal speech output: this model does not generate audio and does not support the Live API. It also does not generate images, so it cannot directly deliver voice-overs, real-time voice conversations, or finished images; use the corresponding generation services for these needs.
The native long-context limit is not a budget that can be fully used on every call. Message history, material content, and output requirements need to be planned together; thinking also consumes Tokens. The calling guide recommends setting max_tokens to 512 or above to avoid having no visible response due to an insufficient budget.
Native tool capabilities do not mean that a single image-and-text request will automatically run code or control a computer. Function calling in Chat Completions requires an execution and result-return process; file and external system operations are also constrained by entry-point capabilities, connection status, and authorization, and cannot be bypassed with prompting alone.

Frequently Asked Questions

Answers to common questions about using gemini-3-flash-preview.

Is Gemini 3 Flash Preview a fixed-date version?

No. Its official code is gemini-3-flash-preview, and it is a preview version with no fixed date in its name. It is also not an alias for Gemini 3.1 Pro or another Flash model. Applications should explicitly store the model ID used and test critical tasks after adjusting versions.

How do I submit an image and get a text response?

Call /gemini/chat/completions, set model to gemini-3-flash-preview, and combine text and image_url in the content array of messages. Images can use an accessible URL or a Base64 data URI; for non-streaming responses, read from message.content in choices.

Can I directly analyze PDFs, videos, and recordings?

The native model supports PDF, video, and audio understanding, but that does not mean both platform endpoints support directly submitting these media. PDFs can be analyzed through the file_url block of /aichat2/conversations together with the file-reading workflow; image_url in /gemini/chat/completions is for images and should not be treated as a general file field. When using the image-and-text endpoint here to analyze video or recordings, you can first extract key frames from the video and transcribe the recording into text, then submit the images and text; this method analyzes the extracted materials and is not equivalent to directly understanding the complete audio or video.

Why are Tokens sometimes consumed without any body text?

It is a reasoning model, and it may consume reasoning Tokens before generating a visible response. If the output budget is too small, there may be no remaining space to write the body text. As recommended in the guide, set max_tokens to 512 or higher, then increase the budget based on task complexity, and check usage and the finish reason.

Which endpoint should I choose for multi-turn conversations and streaming display?

If you want to maintain the history yourself and control image-and-text requests, use Chat Completions and set stream: true to receive incremental content. If you want to save conversations and continue asking questions by id, use /aichat2/conversations; it also provides SSE and NDJSON formats for conveniently displaying responses and tool events.

Model information · Updated: 2026-10-01. For calling parameters and billing rules, see the API and pricing sections.

Use gemini-3-flash-preview for your next task

Start with a clear goal and judge whether it fits your work based on real results.

Start trying it View pricing

Quickstart

Try the API

API Key

Please obtain an API Key first

gemini-3-flash-preview

A multimodal reasoning model balancing interactive efficiency and long-context understanding

Understand visual and textual clues together

Long materials and reasoning capabilities work together

Move from natural language to structured delivery

Screenshot-driven product reviews

Code review and implementation discussions

File Q&A and ongoing analysis

Choose Flash for everyday complex tasks, evaluate Pro for deep challenges

Choose image analysis and generation separately

Prepare tasks and materials

Try it in the API testing area

Integrate according to the API documentation

Use gemini-3-flash-preview for your next task

Quickstart

More from Google