Google Releases Gemini 3.1 Pro With Ability to Execute Complex Tasks; Pomelli Gets New Photoshoot Feature

Google released its most capable Gemini 3 series AI models, dubbed Gemini 3 Pro and Gemini 3 Deep Think, last year. During the release, the Mountain View-based tech giant highlighted that its new AI models outperformed OpenAI's GPT-5.1 on multiple benchmarking platforms. Now, the US-based online search giant has released another AI model as part of the Gemini 3 series. Called Gemini 3.1 Pro, the new AI model builds on the capabilities and functionalities of the Gemini 3 Pro. The company says that Gemini 3.1 Pro is designed to execute more complex tasks. Separately, Google Labs, the company's testing bed for AI experiments, has introduced a new feature for Pomelli, its AI-powered marketing assistant.

Gemini 3.1 Pro Availability, Features

On Thursday, the Mountain View-based tech conglomerate announced that its new Gemini 3.1 Pro AI model, part of its flagship Gemini 3 series, is now rolling out to developers, enterprises, and individual users. The company said that Gemini 3.1 Pro will be available for developers in preview via the Gemini API in Google AI Studio, Gemini CLI, Android Studio, and Google Antigravity, the firm's agentic development platform.

Meanwhile, Google is rolling out the new Gemini 3.1 Pro AI model for enterprises via Vertex AI and Gemini Enterprise products. Lastly, individual users can access the latest AI model through the Gemini app and NotebookLM. The tech giant also highlighted that Gemini 3.1 Pro is rolling out with higher limits for users with Google AI Pro and Ultra plans. Similarly, Gemini 3.1 Pro in NotebookLM is exclusively available to Google AI Pro and Ultra subscribers.

Coming to its performance, Google claims that the Gemini 3.1 Pro is more capable of capable at processing complex problem-solving queries. The AI model achieved a “verified” score of 77.1 percent on the ARC-AGI-2 benchmark, which evaluates the ability of models to solve new logic patterns. Meanwhile, Gemini 3 Pro got a score of 31.1 percent, OpenAI's GPT-5.2 achieved 52.9 percent, and Anthropic's Opus 4.6 got a score of 68.8 percent.

On Humanity's Last Exam test, which evaluates academic reasoning capabilities, Gemini 3.1 Pro scored 44.4 percent, outperforming Gemini 3 Pro and OpenAI's GPT 5.2, which scored 37.5 percent and 34.5 percent, respectively.

Google demonstrated that the Gemini 3.1 Pro AI model can generate significantly better Scalable Vector Graphics (SVG) animations than the Gemini 3 Pro, which was launched in November 2025. Both models were given the same prompts. However, the latest model was able to add greater details to the SVG.

Google's Pomelli Can Now Create Image Portfolios

Separately, the Mountain View-based tech giant's Google Labs, an incubator for experimental AI features and tools, has announced that it is introducing a new feature for Pomelli. The AI tool for marketing now gets the Photoshoot feature, allowing users to create a portfolio of multiple images and banners for campaigns from a single reference image of any product.

For example, if a user uploads an image of a phone and asks Photoshoot in Pomelli to generate images for commercials, it will create multiple product shots from various angles with promotional text. In terms of availability, the feature is currently rolling out for free to users in the US, Canada, Australia, and New Zealand.