Models
Gemini (3.1 Pro)
Google DeepMind's natively multimodal model family, notable for very long context windows and strong reasoning benchmarks.
Model Facts
Provider
Google DeepMind
Released
March 1, 2026
Context window
1M tokens
Weights
Closed
GPQA Diamond
94.3%
ARC-AGI-2
77.1%
What it is
Gemini is Google DeepMind's model family, powering the Gemini app, Google Search's AI features, and Vertex AI. It was designed from the start to be natively multimodal — trained jointly on text, images, audio, and video rather than having those capabilities bolted on afterward.
Where these numbers come from
The benchmarks above are for Gemini 3.1 Pro, which scores 94.3% on GPQA Diamond and 77.1% on ARC-AGI-2 (it's also been reported at 44.4% on Humanity's Last Exam, one of the harder general knowledge/reasoning benchmarks in current use). Google has a Gemini 3.5 Pro release clearing for launch around the same time this entry was written — update this entry via the admin dashboard once that version is generally available with confirmed official benchmarks.
Notable traits
- Consistently among the largest context windows offered by any mainstream closed model — Gemini hit 1 million tokens well before most competitors and has held that lead across generations.
- Built on a Mixture-of-Experts architecture, which activates only a subset of the model's parameters for any given input rather than the whole network.
- Closed weights — available through Google AI Studio, the Gemini API, and Vertex AI.