Gemini 3.1 Pro is Now Live on CometAPI: What it is and how to access

The Gemini 3.1 Pro is now available on CometAPI, and you can start using it through CometAPI's services—at a more affordable launch price than the official price. CometAPI already exposes the Gemini 3 family and provides an OpenAI-compatible path to call those models from a single unified gateway; that makes it quick to experiment with Gemini models using existing OpenAI SDKs

What is Gemini 3.1 Pro? (Why does this matter?)

Gemini 3.1 Pro refers to a point-release / preview iteration in the Gemini-3 family — the same flagship model family from Google that powers Gemini 3 Pro and related "Flash" variants. Gemini 3.x builds focus on deeper multimodal reasoning (text + images + video), higher token/context windows, improved code generation ("vibe coding" and agent workflows), and incremental performance/efficiency improvements over the initial Gemini 3 Pro release. The Gemini 3 family is now widely available across Google’s APIs and cloud offerings.

Key technical highlights (what the model delivers)

• Multimodal inputs: text + images + (in some variants) video/audio.
• Very long context windows (document- and code-scale interactions up to ~1M tokens in supported variants).
• Agentic tooling and coding improvements — designed for autonomous agents, IDE integrations, and complex code tasks.

Benchmark of Gemini 3.1 Pro

The benchmark results for the Gemini 3.1 Pro are as follows: AIME 2025: 100% (including code execution); SWE-Bench Verified: 83.9%; ARC-AGI-2: 71.8%; LiveCodeBench Pro: 2844; Elo Terminal-Bench 2.0: 63.5%; MMMLU: 93.6%. For reference, the Gemini 3 Pro scored 76.2% on SWE-Bench. This will be a huge leap forward. 84% on SWE is huge.

vs Gemini 3.0 pro：

Across the 23 benchmarks provided, Gemini 3.1 Pro shows an average improvement of approximately 17.5%.

However, as the screenshots illustrate, the real "differences" lie in the following areas:

The most significant improvements are not in "fact-taking," but rather in complex reasoning and reliability in long contexts: ARC-AGI-2 (+130.9%): This is the most significant improvement. ARC measures the model's ability to learn new skills (fluid intelligence) on the fly, rather than relying on training data. The increase from 31.1% to 71.8% indicates a significant architectural shift in how the model "thinks."

Vending-Bench 2 (+52.3%): This benchmark tracks the "net worth" of an agent in its environment. The leap here demonstrates that version 3.1 significantly outperforms other versions in planning, executing multi-step tasks, and effectively managing "tools."

Long context (MRCR v2 1 million points): While the standard context (128,000 tokens) offers a 10% performance improvement, the leap of up to 51.3% at 1 million tokens demonstrates that version 3.1 is more stable and accurate when handling massive amounts of data.

Is CometAPI offering Gemini 3 / 3 Pro / 3.1?

Yes — CometAPI publicly lists the Gemini 3 family (Gemini 3.1 Pro entries and related model strings) in its model catalog and marketing copy, and shows an OpenAI-compatible integration path (single API key + base_url) for calling models. CometAPI presents examples that use an OpenAI client pointed at https://api.cometapi.com/v1 and model names in the model field. That means you can often call Google Gemini family models through CometAPI with the familiar OpenAI-style client code. Gemini 3.1 pro is now offered through CometAPI’s unified model marketplace and playground.

Step-by-step: quick checklist to get started

1. Create a CometAPI account and get an API key (CometAPI offers a free key / trial in many cases).
2. From CometAPI’s console or model catalog, confirm the exact model name (gemini-3.1-pro-preview).
3. Use CometAPI’s OpenAI-compatible endpoint (base_url) and set the model parameter to the model alias from the catalog.
4. Start with test prompts, check latency/cost, then scale with rate limits and engineering controls (timeouts, retry logic, streaming, etc.).

Before accessing, please make sure you have logged in to CometAPI and obtained the API key. CometAPI offer a price far lower than the official price to help you integrate.

Ready to Go?→ Sign up fo Gemini 3.1 pro today !

If you want to know more tips, guides and news on AI follow us on VK, X and Discord!