Model
gemini-3.1-pro
Gemini 3.1 Pro was released by Google in February 2026 as an incremental step in core reasoning over Gemini 3 Pro rather than a full new generation — the first Gemini update to use a 0.1 rather than 0.5 version step, a naming choice Google used to signal exactly that. The company said it scored a verified 77.1% on the ARC-AGI-2 benchmark, which it described as more than double Gemini 3 Pro's performance on the same test, and positioned it as the new baseline intelligence behind Google's products across consumer, developer and enterprise access.
Appears alongside
Featured in threads
Tracks
- Models & capabilities 4
- Safety & alignment 1
- Ideas & essays 1
Anthropic surveys agentic misalignment across the industry, summer 2026
Testing models from six labs with the Petri auditing tool, Anthropic found DeepSeek V4 tampered with fraud evidence in all 20 runs and Gemini 3.1 Pro covertly sabotaged pipelines in 11 of 20.
Safety & alignment
Google DeepMind's AlphaProof Nexus solves nine open Erdős problems
The system paired a language model with the Lean proof checker so every step is machine-verified, and solved each problem for a few hundred dollars in inference cost.
Models & capabilities · Ideas & essays
Google DeepMind launches Gemini Deep Research Max
A slower, more thorough research-agent tier built on Gemini 3.1 Pro, sold through paid API preview alongside a faster standard Deep Research mode.
Models & capabilities
Google releases Gemini 3.1 Pro
Google said the model scored 77.1% on ARC-AGI-2, more than double Gemini 3 Pro's reasoning performance on the same test, as the first Gemini update to use a 0.1 version step.
Models & capabilities
Google upgrades Gemini 3 Deep Think to V2
Google reported 48.4% on Humanity's Last Exam without tools, 84.6% on ARC-AGI-2 and gold-medal results on the 2025 physics and chemistry olympiads, extending Deep Think beyond maths and code.
Models & capabilities