Gemini 3 Deep Think goes live for Ultra subscribers with early API access
Google's most capable reasoning model is now available to Gemini Ultra subscribers, with API access rolling out to researchers and enterprises. Positioned for hard technical problems — science, engineering, and multi-step analysis — not casual chat.
Original sourceGoogle just made its biggest play in the reasoning model race. Gemini 3 Deep Think isn't another chatbot upgrade — it's a dedicated reasoning engine built for problems that require sustained, multi-step thinking.
The model is live today in the Gemini app for Ultra subscribers ($20/mo), with API access opening to researchers, engineers, and enterprise customers over the coming weeks.
What makes Deep Think different from standard Gemini? It uses extended "thinking time" — similar to OpenAI's o-series approach — where the model reasons through complex problems before responding. Google claims significant improvements on math, coding, and scientific benchmarks.
The positioning is deliberate: this isn't for writing emails or summarizing articles. Google is targeting the hardest problems — drug discovery, chip design, theorem proving, and complex engineering calculations.
For developers, the API pricing hasn't been announced yet, but early access is available through Google AI Studio. The model supports the same function calling and multimodal inputs as other Gemini models.
Panel Takes
The Builder
Developer Perspective
“Finally, a reasoning model with multimodal support. Being able to feed it diagrams, code screenshots, and technical documents while it reasons through problems is a genuine unlock. The API can't come fast enough.”
The Skeptic
Reality Check
“Google keeps shipping models that benchmark well but feel different in practice. I'll reserve judgment until I've actually used it on real engineering problems. The o-series set a high bar.”
The Futurist
Big Picture
“This is the beginning of specialized AI models for specific cognitive tasks. We're moving from 'one model does everything' to a toolkit of reasoning specialists. The implications for scientific research alone are enormous.”