Google has unveiled an updated preview of its Gemini 2.5 Pro AI model, promising significant improvements in coding capabilities and performance across a range of programming benchmarks. The latest version exhibits enhanced math, science, and reasoning abilities, featuring better response formatting and creativity. More notably, the model now supports native multimodal input across text, audio, images, and video, facilitating a more comprehensive approach to complex problem-solving.
A standout feature of Gemini 2.5 Pro is its Deep Think mode, which augments the model’s reasoning capability, making it particularly adept at tackling mathematical reasoning and coding tasks. This advanced technology positions Gemini 2.5 Pro ahead of competitors like OpenAI’s o3, Claude Opus 4, DeepSeek R1, and Grok 3 Beta, with marked improvements demonstrated through benchmarks such as AIME mathematics, LiveCodeBench coding, SWE-bench software engineering tasks, LMArena, and WebDevArena coding performance benchmarks.
With a 1-million token context window, Gemini 2.5 Pro provides extensive context for inputs, enabling it to outperform rival models. Priced competitively at $1.25 per million input tokens and $10 for output tokens, it offers a robust option for enterprises seeking advanced AI solutions. The model is currently available for enterprise testing through Google AI Studio, Vertex AI, and the Gemini app, with general availability expected shortly.

Collections

Google released an updated preview of Gemini 2.5 Pro with significant improvements in coding, math, and reasoning capabilities. The model features native multimodal support for text, audio, images, and video, plus a new Deep Think mode for enhanced reasoning. With a 1-million token context window and competitive pricing at $1.25 per million input tokens, it outperforms competitors like OpenAI's o3 and Claude Opus 4 on various benchmarks. Currently available for enterprise testing through Google AI Studio and Vertex AI.

Gemini 2.5 Pro: Google's Enhanced AI Model for Coding and Problem-Solving