Gemini 3.1 Pro — reason through code and documents
Gemini 3.1 Pro brings text, images, audio, video and PDFs into one conversation. Choose a thinking level, request structured text, and keep the original conversation parts when your application calls tools. $0.80 / 1M input · $4.80 / 1M output.
- 1M-token input
- 65,536-token output
- Three thinking levels
- Function calling
- Multimodal input
Playground
Explore Gemini 3.1 Pro with native requests. Choose an account group, inspect the settings, and submit using your balance. Live capability verification is pending.
gemini-3.1-pro-preview$0.80 in · $4.80 out per 1M tokensMore Gemini parameters
Try one of these prompts
Messages are billed from your balance at the rates below, like an API call. Failed requests are free.
Gemini 3.1 Pro pricing
Gemini 3.1 Pro prices below come from current billing settings. Compare available account groups, including their input, output and cached-input rates. Requests above 200,000 uncached input tokens use the long-context tier.
Example sizes are for illustration. Your bill follows the tokens each request reports. Standard processing rates. The 200,000-token boundary counts uncached input. Cached tokens are billed separately and are not billed again as uncached input. Tool charges and cache storage are separate from these token examples.
| Category | googlellm · ≤ 200,000 uncached input tokensPrice · 1M tokens | googlellm · > 200,000 uncached input tokensPrice · 1M tokens | googlellm-vertext · ≤ 200,000 uncached input tokensPrice · 1M tokens | googlellm-vertext · > 200,000 uncached input tokensPrice · 1M tokens |
|---|---|---|---|---|
| Input | $0.80 | $1.60 | $1.00 | $2.00 |
| Output, including thinking | $4.80 | $7.20 | $6.00 | $9.00 |
| Cached input | $0.08 | $0.16 | $0.10 | $0.20 |
Only uncached input determines the tier. Cached input is billed separately at the rate for that tier. Examples use the standard tier.
What is Gemini 3.1 Pro?
Gemini 3.1 Pro is a Google reasoning model for understanding code, comparing documents and explaining information across media. Supply the material, question and expected answer format together.
Use gemini-3.1-pro-preview for regular Gemini 3.1 Pro. Google also documents a customtools variant for specialized tool selection. This page connects the regular version; thinking depth is a separate setting.
Official model limits checked October 17, 2026. Input and output have separate limits.
Start with Gemini 3.1 Pro
Start with a representative task. Save its inputs and settings for comparison.
Try the Playground
Send a Gemini 3.1 Pro message, attach supporting material, and inspect the native JSON request before submitting. Official controls use native field names and types; custom JSON is sent unchanged for the Gemini API to validate.
Choose your account group
Choose among the Gemini 3.1 Pro groups available to your account. The Playground updates rates for your selection; additional groups appear through the same account-aware list.
Use the API reference
Copy a native Gemini 3.1 Pro example into your application. The reference lists request fields, response fields, media inputs and capabilities that still require a live check.
What Gemini 3.1 Pro can help you build
Choose an output contract, then check the result against your source material.
Reason through a change
Ask Gemini 3.1 Pro to trace a bug across related functions, explain its assumptions, and propose a focused patch. Run your tests before accepting the change.
Return structured data
Give Gemini 3.1 Pro a JSON response schema for extraction or classification. Check the returned values against the document; valid JSON alone does not establish factual accuracy.
Work with your functions
Declare named functions and their argument schemas. Gemini 3.1 Pro can return a function call for your application to execute; your code decides which actions are permitted.
Compare media and text
Place a chart, screenshot or PDF beside the question. Gemini 3.1 Pro accepts multimodal input and returns text, making it useful for explanations grounded in supplied material.
Practical Gemini 3.1 Pro tasks
Suggested workflows; no benchmark or customer outcome is claimed.
Review a repository change
Provide the relevant files and test failure to Gemini 3.1 Pro. Ask for the smallest change that addresses the failure, with a separate explanation of remaining uncertainties.
Extract a document checklist
Ask Gemini 3.1 Pro to turn a supplied document into a structured checklist. Include source locations in the schema so a reviewer can check each extracted item.
Explain a recorded demonstration
Provide a short supported video and ask Gemini 3.1 Pro to describe a specific step. Limit the question to visible evidence and ask it to identify unclear details.
How to use Gemini 3.1 Pro
- 01
Prepare access
Sign in, add credit and create a server-side API key. Open Gemini 3.1 Pro in the Playground to inspect the request format.
- 02
Supply the context
Put the prompt and previous conversation in contents. For Gemini 3.1 Pro media input, include the correct MIME type and a supported file reference.
- 03
Choose thinking and output
Use low, medium or high thinking. Gemini 3.1 Pro defaults to high; set maxOutputTokens deliberately when response length matters.
- 04
Check the response
Read candidates, finishReason and usageMetadata. Preserve Gemini 3.1 Pro thought signatures and tool parts when sending the next turn.
The native Gemini 3.1 Pro API
Gemini 3.1 Pro uses generateContent for JSON responses and streamGenerateContent for server-sent events. The separate reference includes copyable examples and the complete parameter schema.
Exact model name
Use gemini-3.1-pro-preview in the URL. A display name such as Gemini 3.1 Pro is not an interchangeable API identifier.
Explicit generation settings
Gemini 3.1 Pro retains temperature and sampling fields. Google recommends temperature 1.0. Thinking level, output format and tools remain distinct controls.
Read actual usage
Gemini 3.1 Pro output usage includes thinking tokens. Cached input stays inside total input for the context threshold and is priced at the cached-input rate.
Two official Gemini 3.1 Pro variants
The distinction is tool selection, not a second reasoning level. Only the regular Gemini 3.1 Pro model is connected on this page.
| Choice | Regular Preview | Customtools Preview |
|---|---|---|
| API ID | gemini-3.1-pro-preview | gemini-3.1-pro-preview-customtools |
| Documented purpose | General reasoning and tool use | Prioritize custom tools when bash is also available |
| Function calling | Supported | Specialized tool-selection behavior |
| Evaluation question | Does it solve the actual task? | Does it choose the intended custom tool? |
Google warns that Customtools can show quality variation outside its intended workflow. This is a documentation comparison, not a new A/B benchmark. Google model documentation
Gemini 3.1 Pro limits
Separate input and output budgets
Gemini 3.1 Pro accepts up to 1,048,576 input tokens and produces up to 65,536 output tokens. Thinking uses the output budget; a short answer can still require thinking tokens.
Text output only
Gemini 3.1 Pro can analyze supported media, but it does not generate images or audio. Use a dedicated generation model for those outputs.
Check advanced capabilities
Gemini 3.1 Pro supports tools in its official model specification. Live checks for media, caching and individual tools are still pending here; accepting a parameter does not prove its behavior.
Cache resources need access
A cachedContent value identifies an existing cache resource. Gemini 3.1 Pro cannot use a cache merely because its name is well formed; resource access and model compatibility still apply.
Use Gemini 3.1 Pro with SeedRouter
One account balance
Use the same SeedRouter balance for Gemini 3.1 Pro and other listed text, image and video models. Keep each model’s request format and constraints.
Prices beside the model
Check Gemini 3.1 Pro token rates on this page and select the account group you intend to use. The price table follows current billing settings.
Pay for successful requests
Failed Gemini 3.1 Pro requests are not charged. Successful requests use reported token usage, including thinking and cached input.
Explore other text models
Compare documented capabilities and current rates across available text models.
Call GPT-6 Astra with the OpenAI SDK. OpenAI's most capable model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.
View pricingCall GPT-6 Sol with the OpenAI SDK. 1.05M-token context, reasoning effort from none to max, live per-token pricing and a playground you can try right now.
View pricingCall GPT-6.1 Sol with the OpenAI SDK: near-Astra coding and agent work at a fifth of the price, 1.05M-token context, 95% off cached input. Try it now.
View pricingCall GPT-6 Luna with the OpenAI SDK. The lowest-cost GPT-6 model, with a 1.05M-token context, live per-token pricing and a playground you can try right now.
View pricingTry Claude Haiku 5.5 for classification, extraction and summaries. See live token prices, test your prompts, and connect your app with the Messages API.
View pricingTry Claude Sonnet 5.5 for code reviews and document questions. See live token prices, test a prompt, and check verified capabilities before connecting your app.
View pricingGemini 3.1 Pro guides
Understand the two official model IDs before choosing settings for your application.
Compare the two official Gemini 3.1 Pro model IDs, what Customtools changes, and when the regular Preview model fits your application.
Compare Gemini 3.8 Flash and Gemini 3.1 Pro by thinking defaults, model status, token limits and evaluation criteria. No universal winner is assumed.
Gemini 3.1 Pro FAQ
Which Gemini 3.1 Pro model ID should I use?+
Gemini 3.1 Pro uses gemini-3.1-pro-preview here. The separate customtools variant prioritizes custom tools when bash is also available.
What is the default Gemini 3.1 Pro thinking level?+
Gemini 3.1 Pro defaults to high; low and medium are supported; minimal is not. Do not combine thinkingLevel with thinkingBudget.
Why does Gemini 3.1 Pro pricing change above 200K input tokens?+
The current rates use two tiers, with the higher tier applying above 200,000 uncached input tokens. Cached input is billed separately at the selected tier’s cached-input rate. Check the live table for each account group.
Can Gemini 3.1 Pro generate images?+
Gemini 3.1 Pro accepts supported media but returns text. Use an image model when you need a generated picture.
Do I need Customtools for Gemini 3.1 Pro function calling?+
Regular Gemini 3.1 Pro supports function calling. Customtools changes tool-selection behavior; it is not required for declaring functions.
Try Gemini 3.1 Pro
Start with a representative prompt, inspect the native request, and check the result against your task.