STORY · MODELLER_
Google launches Gemini 3 Deep Think V2 with breakthrough in reasoning and scientific tasks
Google is rolling out an upgraded version of its deep reasoning mode Gemini 3 Deep Think V2, which achieves 84.6% on the ARC-AGI-2 benchmark and demonstrates olympiad-level performance in physics and chemistry. The mode is now available to Google AI Ultra subscribers and via Vertex AI/Gemini API, with focus on practical applications such as error detection in mathematics, modeling of physical systems, and 3D-print optimization.
WHY IT MATTERS
This marks a significant step toward human-level intelligence on complex reasoning tasks, with benchmark creator François Chollet projecting human-AI parity around 2030. The combination of productized test-time compute and concrete engineering applications makes this operationally relevant beyond research laboratories.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.