Gemini 2.5 Pro touts 1M‑token context

Google’s developer documentation lists Gemini 2.5 Pro with a 1,048,576-token input limit and a 65,536-token output limit, putting the model among the company’s long-context offerings for developers building agents and document-heavy applications. Google DeepMind introduced Gemini 2.5 on March 25, 2025, describing 2.5 Pro as its most advanced model for complex tasks and saying it shipped with a 1 million-token context window. Google Cloud said on April 9, 2025 that the model was available in public preview on Vertex AI, where it paired that context window with reasoning and coding features aimed at enterprise workloads. Google’s current Gemini API model page lists Gemini 2.5 Pro as stable, with a latest update of June 2025. ### Where does the 1 million-token figure come from? Google’s March 25, 2025 launch post said Gemini 2.5 Pro “ships today” with a 1 million-token context window and said 2 million tokens were “coming soon.” The company framed that release as part of a broader push to build “thinking” capabilities into its model family for more complex, context-aware agents. (blog.google) Google’s current Gemini API model page gives the more precise number: 1,048,576 input tokens. That same page says the model accepts audio, images, video, text and PDF inputs, and supports tools including function calling, code execution, file search, URL context and search grounding. ### What does Google say Gemini 2.5 Pro is for? (blog.google) Google’s model page describes Gemini 2.5 Pro as its “state-of-the-art thinking model,” built for complex problems in code, math and STEM, and for analyzing large datasets, codebases and documents using long context. That description aligns with the company’s earlier launch language around reasoning-heavy tasks rather than low-latency serving. (ai.google.dev) Google Cloud’s April 9, 2025 post was more specific about enterprise use. It said the one million-token context window lets Gemini 2.5 Pro perform deep data analysis, extract information from dense documents such as legal contracts or medical records, and handle complex coding tasks by comprehending entire codebases. Box Vice President of AI Product Management Yashodha Bhavnani said in that post that Gemini 2.5 made it possible to envision “more powerful agent systems” in which extracted insights trigger downstream actions across multiple steps. (ai.google.dev) ### How much does Gemini 2.5 Pro cost? Google’s Gemini API pricing page lists Gemini 2.5 Pro at $1.25 per 1 million input tokens for prompts up to 200,000 tokens and $2.50 for prompts above 200,000 tokens. The same page lists output pricing, including thinking tokens, at $10 per 1 million output tokens up to 200,000 tokens and $15 above that threshold. (cloud.google.com) Google’s pricing page also says paid access includes context caching and Batch API support, with Batch API offering a 50% cost reduction. That matters for developers running repeated long-context jobs, where reused context and asynchronous processing can change the bill materially even when the base model price is higher than lighter variants. (ai.google.dev) ### How does that compare with Flash? Google said on June 17, 2025 that Gemini 2.5 Flash pricing was updated to $0.30 per 1 million input tokens and $2.50 per 1 million output tokens. In the same post, Google said 2.5 Flash-Lite was introduced in preview for cost- and latency-sensitive workloads that require less model intelligence. (ai.google.dev) Google Cloud had already described Gemini 2.5 Flash on April 9, 2025 as the company’s “workhorse model” with low latency and cost efficiency, while positioning 2.5 Pro for maximum quality and the most complex tasks. Those official descriptions map the tradeoff directly: Pro for deeper reasoning and long-context analysis, Flash for faster and cheaper serving. (developers.googleblog.com) ### What should builders watch next? Google’s Gemini API model page says Gemini 2.5 Pro’s latest update was June 2025, and Google’s March 2025 launch post had said a 2 million-token context window was coming soon. Google’s current public documentation is the place developers can verify whether model limits, pricing tiers or tool support change again. (ai.google.dev) (cloud.google.com)