Google Supercharges Gemini 3 Deep Think for Heavy-Duty Science
Google's beefed-up reasoning model now tackles complex scientific and engineering problems, with API access rolling out to researchers.
Google just dropped an upgrade to Gemini 3 Deep Think, its most specialized reasoning mode. The update sharpens the model's ability to chew through complex science, research, and engineering challenges.
This isn't your everyday chatbot stuff. Deep Think represents Google's push into advanced reasoning territory—the kind of AI that can actually work through multi-step technical problems rather than just summarizing Wikipedia.
The company is also cracking open access via the Gemini API, though for now it's limited to select researchers. Translation: if you're doing serious scientific work and want to kick the tires, you might be able to get your hands on it.
No word yet on broader availability or specific benchmarks showing how much better the updated model performs. Classic Google—ship first, flex the numbers later.