Google announces Gemini 3.1 Pro, says it's better at complex problem-solving
SMRTR summary
Google launched Gemini 3.1 Pro with enhanced problem-solving abilities, scoring a record 44.4% on Humanity's Last Exam and dramatically improving to 77.1% on ARC-AGI-2 logic tests compared to its predecessor's 31.1%. However, the model still trails behind competitors like Claude Opus 4.6 on popular Arena leaderboards.
SMRTR provides this summary for quick context. The original article belongs to Ars Technica.
Read the original article