Google DeepMind Unveils Gemini 3.1 Pro: A New Contender Reshaping the AI Race

February 20, 2026Artificial Intelligence
Google DeepMind Unveils Gemini 3.1 Pro: A New Contender Reshaping the AI Race

Google DeepMind launched Gemini 3.1 Pro on February 19, 2026, marking its most capable AI model for complex tasks to date. Built as the next iteration in the Gemini 3 series, the model processes text, images, audio, video, and entire code repositories through its natively multimodal architecture. With a context window of up to 1 million tokens and 64,000-token output capacity, the model is available through Vertex AI, Google AI Studio, Google Antigravity, and Gemini CLI.


The benchmark results paint a compelling picture of the model's competitive positioning. On ARC-AGI-2 abstract reasoning puzzles, Gemini 3.1 Pro achieved 77.1%, dramatically outperforming Claude Opus 4.6 at 68.8% and GPT-5.2 at 52.9%. In agentic coding on SWE-Bench Verified, the model scored 80.6%, nearly matching Claude Opus 4.6's 80.8%, while in scientific research coding it led the field with 59%. What makes these results particularly noteworthy is the pricing: at $2 per million input tokens and $12 per million output tokens, Gemini 3.1 Pro costs less than half of what Claude Opus 4.6 charges for comparable performance.


Early adopters have reported significant improvements. JetBrains observed up to 15% quality improvement over the best Gemini 3 Pro Preview runs, noting the model delivers more reliable results with fewer output tokens. Cartwheel, a 3D animation company, reported that the model resolved a longstanding rotation order bug in their export pipeline, praising its superior understanding of 3D transformations. Hostinger highlighted the model's ability to translate user intent into production-quality code, describing the reasoning improvement as immediately noticeable.


On the safety front, the model underwent rigorous evaluation under Google's Frontier Safety Framework across five risk domains. While cyber capabilities showed an increase compared to Gemini 3 Pro, the model remained below critical capability levels across all categories. The model has also been integrated into GitHub Copilot, making it available to Pro, Pro+, Business, and Enterprise users. With its combination of strong benchmark performance, competitive pricing, and broad platform availability, Gemini 3.1 Pro represents a significant shift in the AI landscape that competitors will need to address.

📬 Subscribe to Our Newsletter

Stay updated with our latest blog posts and updates.

English

*You can unsubscribe at any time.