Google releases Gemini 2.5 Pro/Flash to the public & newly releases cheap and fastest 'Gemini 2.5 Flash-Lite' preview version
Google has expanded its Gemini 2.5 model family, making Gemini 2.5 Flash and Gemini 2.5 Pro generally available, and also unveiling a preview of Gemini 2.5 Flash-Lite, the most cost-effective and fastest Gemini 2.5 model to date.
Gemini 2.5 model family expansions
Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities.
(PDF file) https://storage.googleapis.com/deepmind-media/gemini/gemini_v2_5_report.pdf
Gemini 2.5 is Google's family of hybrid inference models designed to be at the Pareto-optimal frontier of cost and speed while providing excellent performance. Google is now announcing the general availability of Gemini 2.5 Pro and Gemini 2.5 Flash as stable versions following user feedback.
The Gemini 2.5 Pro, which was previewed alongside the announcement of the Gemini 2.5 family, is the most powerful model developed by Google, capable of high-level coding and inference, as well as multi-modal understanding capable of processing up to three hours of video content.
Google announces next-generation inference AI model 'Gemini 2.5', greatly improving inference and coding performance - GIGAZINE
Gemini 2.5 Pro has been released as a preview version, and an enhanced version, Gemini 2.5 Pro Preview (I/O edition), was released in May 2025. Although the coding performance of this I/O edition has improved, it has been pointed out that other performance has deteriorated, and Google's development team has promised users that it will 'fix it.' This official version reflects this fix.
Google releases early access version of 'Gemini 2.5 Pro Preview (I/O edition)' AI model with enhanced coding capabilities - GIGAZINE
Gemini 2.5 Pro is now officially available on the Gemini app for smartphones. It will also be available for limited access for free plan users. Input price is $1.25 per million tokens and output price is $10.00 per million tokens.
Gemini 2.5 Flash, announced in April 2025, is an inference model with reduced compute and latency requirements, and like Gemini 2.5 Pro, is built as a native multi-modal model supporting long contextual inputs of over 1 million tokens, including text, audio, images, videos, and entire code repositories.
Google announces 'Gemini 2.5 Flash', claiming it is more cost-effective than OpenAI's 'o4-mini' - GIGAZINE
Gemini 2.5 Flash has an input price of $0.30 per million tokens and an output price of $2.50 per million tokens, and is available on Google AI Studio and Vertex AI. Gemini 2.5 Flash can also be accessed through the Gemini app.
Additionally, Google has announced a preview of a new model, Gemini 2.5 Flash-Lite.
Gemini 2.5 Flash-Lite offers higher overall quality than Gemini 2.0 Flash-Lite in coding, math, science, reasoning and multimodal benchmarks. It performs particularly well on high-volume, latency-sensitive tasks like translation and classification, and has lower latency than Gemini 2.0 Flash-Lite and Gemini 2.0 Flash across a broad range of prompt samples. However, Thinking mode is turned off by default and can be turned on via an API parameter.
Gemini 2.5 Flash-Lite, like other Gemini 2.5 models, has the ability to turn on thinking at different budgets, connections to tools like Google search and code execution, multimodal input, and a context length of 1 million tokens. Google said the goal of Gemini 2.5 Flash-Lite is to 'provide an economical model class that offers ultra-low latency capabilities and high throughput per dollar.'
Below is a table summarizing the API usage prices and benchmark results for the Gemini 2.5 family, including Gemini 2.5 Flash-Lite. The input price for Gemini 2.5 Flash-Lite is $0.10 (about 15 yen) per million tokens, and the output price is $0.40 (about 58 yen) per million tokens.
A preview version of Gemini 2.5 Flash-Lite is currently available in Google AI Studio and Vertex AI , alongside the stable versions of Gemini 2.5 Flash and Pro. Google Search will also introduce custom versions of Gemini 2.5 Flash-Lite and Gemini 2.5Flash to its AI-powered overviews and AI mode .
in Software, Posted by log1i_yk