Gemini 3.1 Pro
Multimodal
New
#1ARC-AGI-2
#1Terminal Bench 2.0
#3SWE Bench Verified
+1 more
by Google DeepMind
+
+
+
+
About
Gemini 3.1 Pro is a multimodal language model from Google DeepMind, released in preview in February 2026 as a point-version upgrade to Gemini 3 Pro focused on improving reasoning depth, factual grounding, and coding and agentic task performance. The model accepts text, images, video, audio, and PDFs as inputs across a 1M token context window, extending the multimodal breadth of the Gemini 3 series with a companion endpoint specifically optimized for custom tool use in agentic pipelines. Google describes it as built to refine the reliability and performance of the Gemini 3 Pro series, reflecting an incremental engineering iteration rather than an architectural overhaul.
+
+
+
+
Pricing Range
Input (per 1M)$2.00 -$2.00
Output (per 1M)$12.00 -$12.00
Providers1
+
+
+
+
Timeline
ReleasedFeb 19, 2026
Knowledge CutoffJan 1, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown
Overall Performance
4 benchmarks
Average Score
73.1%
Best Score
80.6%
High Performers (80%+)
1Performance Metrics
Max Context Window
1.1MTop Categories
Coding
77.7%
Reasoning
77.1%
Finance
59.7%
+
+
+
+
All Benchmark Results for Gemini 3.1 Pro
Complete list of benchmark scores with detailed information
| SWE Bench Verified | Coding | 80.60 | 80.6% | Unverified | |
| ARC-AGI-2 | Reasoning | 77.10 | 77.1% | Unverified | |
| Terminal Bench 2.0 | Coding | 74.80 | 74.8% | Unverified | |
| Finance Agent | Finance | 59.72 | 59.7% | Unverified |