Gemini 3.1 Pro hits 77.1% on ARC-AGI-2

// 66d agoMODEL RELEASE

Gemini 3.1 Pro hits 77.1% on ARC-AGI-2

Google has released Gemini 3.1 Pro in preview as its latest reasoning model for complex, agentic workflows. The model supports a 1 million-token context window, 64K-token output, and a verified 77.1% score on ARC-AGI-2.

// ANALYSIS

Gemini 3.1 Pro pushes Google deeper into agentic, tool-using workflows. Its 1M-token context window, 64K output limit, and 77.1% ARC-AGI-2 result make it especially relevant for long-running coding and multi-step tasks.

// TAGS

gemini-3-1-progoogle-deepmindagentreasoning-modelsstitch-design-systemcoding-aimultimodal

DISCOVERED

66d ago

2026-03-23

PUBLISHED

66d ago

2026-03-23

RELEVANCE

10/ 10

AUTHOR

WorldofAI

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE47m ago

Supabase Auth opens Passkeys public beta

Supabase has opened the Passkeys public beta to all projects, enabling passwordless, phishing-resistant logins via biometrics and hardware keys. Built on the WebAuthn standard, the feature supports discoverable credentials for a "username-less" sign-in experience.

INFRA51m ago

Hippocratic AI hits 99.9% safety on NVIDIA Blackwell

Hippocratic AI achieved 99.9% clinical safety and a 2x prefill speedup using DigitalOcean’s NVIDIA Blackwell-powered AI-Native Cloud. The collaboration demonstrates the real-world performance gains of the HGX B300 for high-concurrency, safety-critical medical agents.

NEWS53m ago

Microsoft debuts homegrown AI coding models

Microsoft is unveiling a suite of in-house AI models at next week's Build conference, led by a new coding model designed to power GitHub Copilot and reduce reliance on OpenAI.