CALIFORNIA: Google has released Gemini 3.1 Pro, a new version of its flagship artificial intelligence model that the company says is built to execute complex, multi step work such as synthesizing large amounts of information, following detailed instructions, and completing longer workflows. Google announced the model on Feb. 19 and began rolling it out across consumer products, developer tools, and business offerings, expanding access through the Gemini app, NotebookLM, and Google Cloud platforms.
Gemini 3.1 Pro is designed for advanced reasoning tasks that require sustained context, including reading and summarizing long documents, writing and debugging code, and explaining complicated topics with structured steps. Google DeepMind’s model card describes Gemini 3.1 Pro as natively multimodal, accepting text, images, audio, and video, with a context window of up to one million tokens and text outputs up to 64,000 tokens. Google said the model more than doubles the reasoning performance of its predecessor on a verified ARC AGI 2 benchmark score.
For consumers, Google is enabling Gemini 3.1 Pro in the Gemini app and NotebookLM, its research and note taking product, with access tied to paid plans. Google said higher usage limits are being provided for Google AI Pro and Google AI Ultra subscribers in the Gemini app. NotebookLM access for Gemini 3.1 Pro is limited to those Pro and Ultra tiers, positioning the model as a premium option for users who need longer context handling and more capable multi step assistance.
Product rollout and developer access
Developers are being offered Gemini 3.1 Pro through Google’s Gemini API in Google AI Studio, along with availability in Gemini CLI and integration paths for building and testing within Android Studio. For enterprises, Google is making the model available through Vertex AI and Gemini Enterprise, its business focused suite. Google also listed the model’s support for tool use patterns common in production applications, including function calling and custom tools workflows, aimed at enabling applications that can plan, retrieve information, and complete tasks through connected services.
Pricing for Gemini 3.1 Pro in the Gemini API is listed on Google’s developer documentation as a paid tier model, with no free tier availability for this version. Google lists input pricing at $2 per one million tokens for prompts up to 200,000 tokens and $4 per one million tokens for larger prompts, with output pricing at $12 and $18 per one million tokens respectively, including reasoning tokens. Google’s documentation also lists separate charges for context caching and for grounding with Google Search beyond included allowances.
Performance metrics and model details
Google’s public benchmark reporting for Gemini 3.1 Pro highlights an ARC AGI 2 verified score of 77.1% under a “Thinking High” configuration, compared with a lower reported score for Gemini 3 Pro under a similar setting. Google has also positioned Gemini 3.1 Pro as part of the broader Gemini 3 family, with the developer guide listing the model as preview software and providing its context and output limits alongside other Gemini 3 variants. The model card states that Gemini 3.1 Pro is based on Gemini 3 Pro.
Google DeepMind’s model card also summarizes internal safety evaluation results, stating that Gemini 3.1 Pro outperforms Gemini 3 Pro across safety and tone measures while keeping unjustified refusals low, and that overall safety results were consistent with the original Gemini 3 Pro safety assessment. Google said Gemini 3.1 Pro outputs text and is intended to provide more reliable handling of complex prompts across a wide range of modalities and use cases as it rolls out across products and platforms. – By Content Syndication Services.