Gemini 3.1 Pro

Released 2026-02 · hybrid · 1.0M tokens · 6 benchmarks

Product

Editorial notes

Lanzado 19 febrero 2026. 2x boost de razonamiento sobre Gemini 3 Pro. GPQA Diamond record historico (94.3%).

Spec sheet

Company: Google DeepMind
Country: US
Type: hybrid
Release: 2026-02
Context: 1.0M tokens
Pricing (google): $2/$12/M
Slug: gemini-3-1-pro

Benchmarks (6)

Reasoning 3

GPQA-Diamond

Graduate-level Physics, Chemistry, Biology — PhD-level questions.

94.3
ARC-AGI-2

Updated ARC challenge — novel and tough abstract reasoning.

77.1
Humanitys-Last-Exam

The hardest known benchmark — novel academic problems.

44.4

Coding 2

Agentic 1

BrowseComp

Web browsing comprehensive benchmark.

85.9

Cite this model

BibTeX · APA

BibTeX

@misc{frontier-gemini-3-1-pro,
  title  = {Gemini 3.1 Pro},
  author = {{Google DeepMind}},
  year   = {2026},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-06-17},
  url    = {https://frontierbenchmarks.com/models/gemini-3-1-pro}
}

APA

Google DeepMind (2026). Gemini 3.1 Pro [Large language model]. Frontier Benchmarks AI. Retrieved 2026-06-17, from https://frontierbenchmarks.com/models/gemini-3-1-pro

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.

Battle vs another model ← All models More from Google DeepMind Methodology