Google Documentation Lists Gemini 3.1 Pro Preview Pricing Modes and Token Limits

Google AI for Developers has listed Gemini 3.1 Pro Preview in its Gemini Developer API documentation, including model identifiers, pricing categories, and token limits for teams evaluating the preview model.

Google lists model IDs and API pricing modes

Google AI for Developers’ Gemini API pricing page identifies Gemini 3.1 Pro Preview under the model IDs gemini-3.1-pro-preview and gemini-3.1-pro-preview-customtools. The same Google pricing documentation lists the model across several API pricing modes, including standard, batch, flex, and priority pricing, with rates presented on a per-1-million-token basis.

Those details are relevant for developers planning deployments because API cost can vary depending on both the selected pricing mode and the amount of text or multimodal input sent to the model. Google’s documentation should be treated as the primary reference for current availability and commercial terms, particularly because the model is labeled as a preview.

A third-party catalog page from Price Per Token, titled “Gemini 3.1 Pro Preview API Pricing 2026” and updated June 21, 2026, lists the model at $2.00 per million input tokens and $12.00 per million output tokens. Price Per Token also summarizes a 1.0 million-token context window, speed metrics, and benchmark-related details. Because it is not Google’s official documentation, its pricing figures are best used as a supplementary reference rather than as the final source for budgeting.

Google model page shows a large context window

Google AI for Developers’ model page for Gemini 3.1 Pro Preview identifies the model as gemini-3.1-pro-preview and lists a 1,048,576-token input limit. The same Google page lists a 65,536-token output limit and says the model’s latest update was in February 2026.

A context window of about one million tokens can be significant for applications that need to work with long documents, large codebases, transcripts, research material, or extended tool interactions. In practice, however, a larger context window does not automatically mean lower cost or better performance. Developers still need to account for the number of tokens submitted, the number of tokens generated, latency, reliability, and how well the model follows instructions on their own tasks.

Google’s model page also describes supported input modalities for Gemini 3.1 Pro Preview. As with pricing, developers should rely on Google AI for Developers’ documentation for model limits, supported behavior, and availability, since preview models can change.

Third-party benchmark details need context

Price Per Token’s catalog entry adds comparison-oriented information, including benchmark and speed details. Its excerpt reports a GPQA score of 94.1 for Gemini 3.1 Pro Preview, alongside the pricing summary and context-window information.

Benchmark numbers can help developers decide what to test first, but they are not a substitute for workload-specific evaluation. Scores can depend on test setup, prompting method, sampling settings, evaluation rules, and whether the result reflects an official provider claim or an independent catalog entry. Teams considering the model for production use should run their own tests on representative prompts, documents, tool calls, and safety requirements.

What developers should verify

The most important caveat is the preview label. Preview models often expose new capabilities before a stable release, but they may come with different lifecycle, compatibility, or availability expectations than generally available models. Google’s current documentation is therefore the key source to check before committing to a production budget or integration plan.

For now, the source-backed facts are clear: Google AI for Developers’ pricing page lists Gemini 3.1 Pro Preview and the model IDs gemini-3.1-pro-preview and gemini-3.1-pro-preview-customtools; Google’s model page lists a 1,048,576-token input limit, a 65,536-token output limit, supported input modalities, and a February 2026 latest update; and Price Per Token provides a third-party catalog entry dated June 21, 2026, with pricing and benchmark context.

For teams evaluating the preview, the next step is not just comparing headline token prices. It is measuring total cost and quality across real usage patterns, including long prompts, retrieved context, generated output length, and any tool-related workflows.