Google Multimodal reasoning
Gemini 3.1 Pro Preview
Bring text, media and long documents into one reviewable workflow.
Gemini 3.1 Pro Preview is a Google route for multimodal understanding, coding and long-context analysis. Use it when a task combines documents with images, audio or video and the output needs a coherent written decision. The preview label makes version identity and regression checks part of the integration plan.
- Multimodal input
- Long context
- Preview route
Preview availability and behavior can change. Confirm the exact TokenHot route, supported media types, file limits and tool parameters before committing to a production workflow.
AI Chat
Send a prompt and see the response here.
gemini-3.1-pro-previewStart a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Tier | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|---|
| default | input<20K | $2 | $12 | $0.014 | $4.5 |
| default | input>=27K | $4 | $18 | $0.014 | $4.5 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is Gemini 3.1 Pro Preview?
The TokenHot catalog lists text, image, audio, video and document inputs, text output and a roughly 1.05M-token context value. That makes it useful for evidence-heavy review, but does not make every media format or size available in every request path.
Treat a preview model as a versioned dependency. Keep representative mixed-media fixtures, record the route name and compare extraction accuracy and failure modes after an update.
Keep evidence together
Give each source a role and a date. Ask the model to separate what is visible or heard from its interpretation.
Chunk by decision
Use a clear output schema for long reports. Require references back to source sections rather than a smooth but untraceable summary.
Test before scaling
Pin fixtures and monitor regressions. Keep a fallback route for requests that exceed media or context limits.
From brief to deliverable
Put Gemini 3.1 Pro Preview to work
Product, research and QA teams
Review a product demo from multiple sources
Combine a transcript, a screen recording and a written requirement into one issue list. Ask for timestamps and concrete acceptance checks so a reviewer can replay each finding.
- Tools you need
- A media ingestion path and a player for human verification.
- Keep in mind
- The model's written interpretation is not a substitute for replaying a disputed moment.
View a task brief
Review the supplied demo media, transcript and requirements. Produce a prioritized issue list with timestamps, observed evidence, likely impact and a concrete acceptance check. Mark any segment that cannot be read confidently and do not infer unseen behavior.
Research and strategy teams
Compare a long set of reports
Ask the model to extract only decision-relevant facts from several dated reports, then reconcile disagreements and retain the competing values.
- Tools you need
- Document retrieval or file upload, plus a human review of cited sections.
- Keep in mind
- Long context does not remove source quality problems or make a later report automatically authoritative.
View a task brief
Compare the supplied reports using the requested dimensions. Return agreed facts, conflicts, missing evidence and a recommendation. Cite document names and sections for every factual statement; never silently choose between conflicting sources.
Frontend teams
Turn interface screenshots into implementation notes
Use screenshots, interaction rules and existing code to prepare a focused frontend handoff. Ask for visual observations first, then implementation hypotheses.
- Tools you need
- Image input, code access and a browser for actual interaction checks.
- Keep in mind
- A screenshot cannot establish focus order, network behavior or responsive transitions.
View a task brief
Review the supplied screenshots and code against the stated interaction requirements. List visible mismatches, likely causes and the smallest fixes. Include desktop, keyboard and mobile checks, and label every unverified interaction.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Multimodal input | Text, image, audio, video and document inputs are listed | Use a common task schema for mixed evidence and reject unsupported file types before sending. |
| Context capacity | 1,048,576 tokens shown in the catalog | Long context is useful for comparison, but retrieval, chunking and citation checks still belong in your application. |
| Output contract | Text output on this route | Ask for structured sections, timestamps or citations when downstream code needs stable parsing. |
| Preview status | The model name includes Preview | Record the model ID and fixture set so behavior changes are detectable after a catalog or provider update. |
| Endpoint options | Gemini and OpenAI protocol entries are listed | Choose the endpoint that matches your client and verify media encoding, safety and tool parameters on that route. |
The page reflects TokenHot catalog modalities and route metadata. Preview behavior, file limits and protocol-specific options should be verified against the endpoint you call.
Audience & fit
Where Gemini 3.1 Pro Preview fits
Teams reviewing rich customer evidence
Use the route to bring screenshots, recordings and documents into the same first-pass review while keeping a human in the loop for disputed evidence.
Researchers comparing long material
Define comparison dimensions and source names before sending files. Keep the answer auditable instead of optimizing only for a short summary.
Frontend teams preparing handoffs
Pair images with code and interaction requirements. Use the model for issue framing, then reproduce findings in a browser.
Cost context
Plan the cost of a useful result
Plan media preprocessing
Transcoding, OCR, transcription and file retrieval can add cost before the model request. Track those steps separately.
Keep preview fixtures
Run a small representative set after route changes. Record media dimensions, duration and the exact prompt when comparing outputs.
Choose the endpoint intentionally
A Gemini-native and an OpenAI-compatible route may expose different request shapes or limits. Use the full pricing and API documentation for the path you select.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1beta/models/gemini-3.1-pro-preview:generateContent' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"contents": [
{
"role": "user",
"parts": [
{
"text": "Hello"
}
}
]
}'FAQ
Frequently asked questions
What is Gemini 3.1 Pro Preview best suited for?
Gemini 3.1 Pro Preview is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is Gemini 3.1 Pro Preview priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call Gemini 3.1 Pro Preview?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
What does Preview mean here?
It identifies a preview route whose availability or behavior may change. Keep regression fixtures and confirm the current model ID before a production rollout.
Can I send any video or document size?
Not automatically. The catalog lists input modalities, while file size, duration, encoding and endpoint limits still need to be checked for the exact request.
Does the model return images or videos?
This TokenHot route is cataloged with text output. Use a generation route when the deliverable is an image or video.
How do I keep a long review traceable?
Name each source, require section or timestamp references and make the model list conflicts and unknowns separately from its recommendation.
Get started
Build with Gemini 3.1 Pro Preview
API