Google launches Gemini 2.5 Pro

Google has launched Gemini 2.5 Pro, a new AI model. The model processes complex tasks in coding, science, and math. It uses "thinking" capabilities to make decisions through step-by-step processing. Google states Gemini 2.5 Pro performs well in human preference benchmarks and leads in coding, math, and science evaluations.

The Gemini 2.5 family models, including 2.5 Pro, reason before responding. Google designs these models to manage complex prompts. Google combined an enhanced base model with improved post-training for better performance.

Gemini 2.5 Pro scores well in mathematics and science benchmarks. It demonstrates improved coding capabilities, excelling in web app creation and code editing. The model has achieved a 63.8 percent score on SWE-Bench Verified in agentic coding evaluations.

Google made Gemini 2.5 Pro will be available in Google AI Studio and the Gemini app for Gemini Advanced users. The model will integrate into Vertex AI. Google will announce pricing details for production use. Gemini 2.5 Pro supports a 1 million token context window, with expansion to 2 million tokens planned.