Google Releases Gemini 3.1 Pro with 77.1% on ARC-AGI-2
Overview
Google released Gemini 3.1 Pro on February 19, 2026, a mid-cycle upgrade that significantly boosts the model's core reasoning capabilities. Previously limited to the Gemini 3 Deep Think tier, these advanced reasoning features are now being rolled out more broadly across Google's platforms.
Benchmark Performance
Gemini 3.1 Pro scored 77.1% on ARC-AGI-2, an abstract reasoning benchmark that tests general intelligence and pattern generalization. Google claims this is "more than double the reasoning performance" of the previous Gemini 3 Pro—a substantial leap for a mid-cycle release and one of the highest scores publicly reported on this challenging benchmark.
Key Use Cases
- Visual Explanations: Breaking down complex topics with visual reasoning
- Data Synthesis: Extracting insights from large, unstructured datasets
- Creative Projects: Multimodal generation and ideation support
- Agentic Workflows: Multi-step autonomous task execution
Availability
Gemini 3.1 Pro is rolling out in preview across multiple surfaces:
- Gemini app (all users)
- NotebookLM (Pro and Ultra subscribers)
- Developer tools: Google AI Studio, Vertex AI, Gemini Enterprise, Android Studio, and CLI
General availability will follow once Google validates performance across ambitious agentic workflows.
Versioning Change
This release also marks a departure in Google's versioning conventions. Rather than the previous ".5" mid-year cadence, Google is adopting ".1" increments, suggesting a more rapid and iterative release cycle going forward.
Source: 9to5Google | Google Blog
Related Articles
At Google I/O 2026 on May 19, Google unveiled Gemini 3.5 Flash—which outperforms Gemini 3.1 Pro across all benchmarks at 4× the speed and half the API cost—alongside Gemini Spark, a 24/7 personal AI agent that works in the background and can be reached directly via Gmail. Spark enters beta for Google AI Ultra subscribers in the US starting the week of May 26.
Google launched Gemini 3.5 Flash at I/O 2026 on May 19, making it generally available the same day. It outperforms Gemini 3.1 Pro on coding and agentic benchmarks while running 4x faster at 40% lower cost.
Google DeepMind released Gemini 3.1 Pro on February 19, achieving 77.1% on ARC-AGI-2—more than double its predecessor's 31.1%—with a 1M-token context window and 80.6% on SWE-Bench Verified.