Gemini 3.1 Pro
Parameters
Undisclosed
Context Length
1M
Modality
Multimodal
Architecture
Undisclosed
License
Proprietary
Release Date
19 Feb 2026
Knowledge Cutoff
Jan 2025
API Pricing (per 1M)
Input: $2.00 · Output: $12.00
Evaluation Benchmarks
Rank
#49
| Benchmark | Score | Rank |
|---|---|---|
Graduate-Level QA | 0.943 | 🥉 3 |
General Text | 1487 | 12 |
Data Analysis | high 0.79 | 13 |
LiveBench Average | high 0.77 | 16 |
General | high 0.77 | 17 |
Mathematics | high 0.91 | 21 |
Reasoning | high 0.84 | 29 |
Coding | high 0.76 | 32 |
Agentic Coding | high 0.44 | 37 |
Agent Arena | -5.6 | 41 |
Web Development | 1447 | 46 |
Coding Index | 0.69 | 46 |
Intelligence Index | 0.30 | 70 |
Agentic Index | 0.10 | 88 |
Rankings
Overall Rank
#49
Coding Rank
#60
About Gemini 3.1 Pro
Gemini 3.1 Pro is Google DeepMind's flagship multimodal reasoning engine engineered for intricate logic, repository audits, and agentic workflows. It provides native multimodal understanding and granular reasoning control across a 1M token context window.
Technical Specifications
Architecture specifications are undisclosed for proprietary models.
Attention
Attention Structure
Multi-Head Attention
Attention Heads
Undisclosed
Key-Value Heads
Undisclosed
Attention Head Dimension
Undisclosed
Position Embedding
Absolute Position Embedding
RoPE Theta
Undisclosed
Sliding Window Attention
Undisclosed
Sliding Window Size
Undisclosed
Sliding Window Ratio
Undisclosed
Linear Attention
Undisclosed
Linear Attention Ratio
Undisclosed
Normalization
Undisclosed
Activation Function
Undisclosed
Dimensions
Hidden Dimension Size
Undisclosed
Number of Layers
Undisclosed
FFN Intermediate Size (Dense)
Undisclosed
Multi-Token Prediction Heads
Undisclosed
Tokenizer
Vocabulary Size
Undisclosed
Resources
About Gemini 3
Google's latest generation multimodal models with breakthrough performance across coding, mathematics, reasoning, and language understanding. Features ultra-large context windows, native multimodal processing, and thinking modes with minimal latency overhead. Available in Pro and Flash variants optimized for different workloads, with preview versions showing state-of-the-art results on multiple benchmarks.