Source of this article and featured image is TechCrunch. Description and key fact are generated by Codevision AI system.
Google has launched Gemini 3, its most advanced foundation model yet, now accessible via the Gemini app and AI search. The model outperforms previous versions and competitors like GPT-5 Pro on key benchmarks. A specialized variant, Gemini 3 Deepthink, will soon be available to premium subscribers. Experts highlight its enhanced reasoning capabilities and nuanced responses. The release coincides with rapid advancements in AI models from competitors like OpenAI and Anthropic.
Key facts
- Gemini 3 achieved a record 37.4 score on the Humanity’s Last Exam benchmark, surpassing GPT-5 Pro’s 31.64.
- The Gemini app now serves over 650 million monthly active users, with 13 million developers integrating it into their workflows.
- Google introduced Antigravity, a coding interface combining chat, terminal, and browser tools for multi-pane development.
- A research-focused Deepthink variant will be rolled out to Google AI Ultra subscribers after safety testing.
- Gemini 3’s advanced reasoning capabilities were praised by Google’s product head as a significant leap in AI performance.
