TLDR: Google has significantly expanded access to its advanced Gemini 2.5 Pro AI model, making it available to a wider range of developers and even free users. The company has also introduced a competitive two-tier pricing structure for its API, aiming to undercut rivals and accelerate AI development. Gemini 2.5 Pro is lauded for its enhanced reasoning, multimodal capabilities, and strong performance in coding and scientific benchmarks.
Google has announced a major expansion of access to its powerful Gemini 2.5 Pro artificial intelligence model, moving it from an experimental phase into a public preview with increased rate limits for developers. This strategic move, which includes a new, highly competitive pricing model, is designed to attract a broader developer base and solidify Google’s position in the rapidly evolving generative AI market.
Previously, the experimental version of Gemini 2.5 Pro was largely reserved for subscribers, but it is now accessible to free users via the Gemini app, web platform, and Google AI Studio, albeit with stricter usage limits. For developers looking to scale their applications, Google has introduced a structured two-tier pricing model. For usage up to 200,000 tokens per month, developers will pay $1.25 per million input tokens and $10 per million output tokens. Beyond this threshold, prices increase to $2.50 per million input tokens and $15 per million output tokens. These tokens encompass multimodal inputs, including text, images, audio, and video. Google emphasizes that these prices are ‘well below competitors,’ citing Anthropic’s Claude 3.7 Sonnet at $3 per million input and $15 per million output, and OpenAI’s GPT-4o starting at $15 per million input and $60 per million output for outputs. A free tier also offers limited access and free grounding with Google Search for up to 500 queries daily, with additional free queries available in the paid tier before a per-query charge applies.
Gemini 2.5 Pro is touted by Google as its ‘most intelligent AI model to date,’ showcasing impressive capabilities across various domains. It excels in handling complex prompts, advanced reasoning, and intricate coding tasks. The model has demonstrated remarkable performance in scientific benchmarks, achieving 84% accuracy on the GPQA Diamond scientific benchmark test, surpassing typical human expert scores of 70%. Its multimodal nature allows it to process diverse data types seamlessly, supported by an extended context window of 1 million tokens, with plans to expand to 2 million. This enables the AI to comprehend and interact with large documents, audio, and video files more effectively.
Also Read:
- Google Launches “Gemini for Government” for US Federal Agencies at Nominal Annual Fee
- Google Cloud Unveils Agent Development Kit for Real-Time Voice Agents Powered by Gemini 2.0
Developer feedback has been overwhelmingly positive, with Google reporting an over 80% increase in demand across Google AI Studio and the Gemini API. Developers commend its effectiveness and cost efficiency, particularly for programming and coding applications. Significant improvements have been noted in front-end and UI development, code transformations, detailed code editing, and the creation of agentic workflows. The model has even secured the top spot on the WebDev Arena leaderboard, outperforming its predecessor by a substantial +147 Elo points. Silas Alberti from Cognition’s founding team remarked that the updated Gemini 2.5 Pro ‘achieves leading performance’ on junior-dev evaluations and feels ‘like a more senior developer’ due to its ability to make correct judgment calls and choose good abstractions. Michael Truell, CEO of Cursor, also expressed excitement, stating that the latest Gemini 2.5 Pro ‘builds on its already strong real-world coding capabilities.’ The model’s ability to generate a working video game from a single line of text in a demo further highlights its advanced creative and problem-solving prowess. While the model is not yet available on Vertex AI, Google has confirmed its imminent addition.
