Gemini 3.1 Pro Preview

Gemini 3.1 Pro Preview is presented by Google as a Gemini API model, giving developers a preview entry point for applications that can use a current general-purpose model through Google’s API documentation. Its identity as a preview is the clearest documented positioning, but the supplied material does not provide enough technical detail to characterize its architecture, training lineage, or benchmark performance.

The same model is also listed in Google Cloud’s Agent Platform Model Garden, suggesting a practical fit for teams evaluating it in a managed cloud environment alongside their application workflows. The available evidence supports its publication and managed listing, but does not establish specific strengths, quality advantages, or comparative claims; those questions should be assessed against fuller model documentation and application-specific evaluation.

OpenCodegemini-3.1-progemini-pro

Open provider page Provider docs

Quick Info

Provider: OpenCode
Model key: gemini-3.1-pro
Release date: Feb 19, 2026
Last updated: Feb 19, 2026
Knowledge cutoff: 2025-01
AI SDK package: @ai-sdk/google
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.00
Output token cost: $12.00

Limits

Output tokens: 65,536 tokens
Context window: 1,048,576 tokens

Transparent token rates

Compare Gemini 3.1 Pro Preview pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Gemini 3.1 Pro Preview

OpenCode Zen

CoveragePreview

Vals.ai's model page for Gemini 3.1 Pro Preview (02/26) confirms the model was released by Google on February 19, 2026, with a 1 million token context window, 66k max output tokens, and pricing of $2.00 per 1M input tokens and $12.00 per 1M output tokens, matching Google's published rates. The default configuration use The page catalogs an extensive set of independent benchmark runs, reporting a Vals Index accuracy of 41.90% (±1.17) at roughly $1.945 per test with about 9 minutes 20 seconds of latency, alongside scores across Code Migration, CyberBench, GPQA Diamond, LiveCodeBench, SWE-bench, Terminal-Bench 2.1, and several domain-sp