Back in November, Google announced the Gemini 3 Pro AI model with state-of-the-art reasoning and multimodal capabilities. Today, Google announced Gemini 3.1 Pro, the successor to the Gemini 3 Pro model, featuring significant improvements over its predecessor and state-of-the-art results on several key AI benchmarks.
Google claims that the 3.1 Pro model is a smarter, more capable model for complex problem-solving. For example, on the ARC-AGI-2 benchmark, which evaluates a model’s ability to solve entirely new logic patterns, this new model achieved a score of 77.1%. Google also claims that this new model now offers clearer visual explanations of complex topics, making it more suitable for sophisticated subjects.
📣 Gemini 3.1 Pro is here!
— Joana Carrasqueira (@jocarrasqueira) February 19, 2026
Overall reasoning improved but on ARC-AGI-2 – which tests for novel logic patterns – it more than doubles 3 Pro’s score.
You can visualize complex topics, organize scattered data, and bring creative projects to life! 🔥 pic.twitter.com/CSrmxYgZzK
As you can notice in the benchmarks below, 3.1 Pro scores a record 94.3% on the GPQA Diamond scientific knowledge benchmark. On the APEX-Agents long-horizon professional tasks benchmark, this model scored 33.5%—a significant jump compared to other leading models in the market.
Find the full benchmark comparison between Gemini 3.1 Pro, Gemini 3 Pro, GPT 5.2, Claude Sonnet 4.6, Opus 4.6, and GPT-5.3-Codex below.
Gemini 3.1 Pro preview is now available for both consumers and developers. For consumers, the Gemini 3.1 Pro model is available in the Gemini app. For Google AI Pro and Google AI Ultra subscribers, the model will be available with higher limits. This new model will also be available on NotebookLM exclusively for Pro and Ultra users.
For developers and enterprise customers, the Gemini 3.1 Pro preview is now available in the Gemini API via AI Studio, Antigravity, Vertex AI, Gemini Enterprise, Gemini CLI, and Android Studio.