Skip to content
AImpact
IT EN
High Foundation Models · 1 min read

Gemini 3.1 Pro: Google's first '.1' update and the ARC-AGI-2 leap

In one sentence Google releases Gemini 3.1 Pro in preview: a verified 77.1% on ARC-AGI-2, more than double Gemini 3 Pro (31.1%), 80.6% on SWE-Bench Verified and 94.3% on GPQA Diamond. API pricing: $2 per million input tokens and $12 output up to 200K tokens.

Verified Official source
ShareLinkedInX
Reading level

Gemini 3 Pro had shipped in November 2025. Three months later, on February 19, 2026, Google releases an intermediate update in preview, the first with a ".1" version number in the Gemini line, and makes a notable leap on one specific benchmark: ARC-AGI-2.

ARC-AGI-2 is a test of novel logic puzzles, designed to measure whether a model can work out a rule it has never seen rather than recognise things it has already met. Gemini 3 Pro scored 31.1%; Gemini 3.1 Pro reaches a verified 77.1%: more than double.

On the other tests Google published it scores 80.6% on SWE-Bench Verified (fixing real bugs in code) and 94.3% on GPQA Diamond (PhD-level science questions). API pricing is $2 per million input tokens and $12 per million output tokens, for requests up to 200,000 tokens.

Companies

Google, Google DeepMind

Tools

Gemini 3.1 Pro

Tags

GoogleDeepMindGeminiGemini 3.1 ProReasoningARC-AGI

Sources