Gemini 2.5 Deep Think launches: How it fares against rivals
Google just dropped Gemini 2.5 Deep Think, its most advanced AI yet.
For $250/month (with a Google AI Ultra subscription), you get access to an AI that uses multiple agents to brainstorm, edit, and blend ideas for sharper results.
You can try it right now in the Gemini app—this launch follows months of feedback and improvements since its first preview at Google I/O.
How is Deep Think different from other AIs?
Gemini 2.5 Deep Think is built to think more like us—it juggles ideas across different agents at once, using reinforcement learning to tackle complex problems step by step.
It's especially geared toward coding, web development, and scientific research where deep reasoning matters.
It outperformed ChatGPT and Grok in tests
In tough tests like Humanity's Last Exam (2,500 questions!), Gemini 2.5 scored 34.8%, beating OpenAI's o3 (20.3%) and Grok 4 (25.4%).
A research version even cracked five out of six International Math Olympiad problems—though it took hours!
The commercial version isn't quite that intense but still brings strong problem-solving skills for everyday users with a subscription.