Gemini 2.5 Pro, Flash Now Generally Available; Flash-Lite Launches in Preview

Google has announced the latest updates to its Gemini 2.5 model family, including the launch of a new lightweight option, revised pricing for its Flash model, and the formal release of its most advanced Pro model.

At the center of the update is Gemini 2.5 Flash-Lite, now available in preview. This new version is designed to be the fastest and most cost-effective option in the Gemini 2.5 lineup, making it well-suited for high-volume tasks such as classification and summarization.

Gemini 2.5 models are part of Google’s “thinking model” framework—AI systems designed to reason through responses before generating them. Each model includes a configurable thinking budget, allowing developers to decide how much cognitive effort the model applies to a given task. This flexibility improves performance and accuracy, and gives developers more control over speed, cost, and output quality.

Introducing Gemini 2.5 Flash-Lite

Gemini 2.5 Flash-Lite offers:

• Lowest latency and cost among 2.5 models

• Improved performance over previous Flash versions (1.5 and 2.0)

• Faster generation with lower time to first token and higher decode speeds

• Optimized for high-throughput tasks like classification or summarization at scale

The model is also a reasoning model, meaning it can dynamically control how much it "thinks" before responding. Unlike other Gemini 2.5 models, Flash-Lite has “thinking” turned off by default to prioritize speed and efficiency. Developers can still enable thinking when needed via an API parameter.

Flash-Lite supports all of Google’s native developer tools, including:

• Grounding with Google Search

• Code Execution

• URL Context

• Function Calling

Pricing Changes to Gemini 2.5 Flash

Google also announced a pricing overhaul for Gemini 2.5 Flash, the mid-tier model now generally available in its final form. The stable release is identical to the 05-20 preview shared during Google I/O.

Key pricing changes include:

• Input tokens: $0.30 per 1M (up from $0.15)

• Output tokens: $2.50 per 1M (down from $3.50)

• Removal of separate pricing for “thinking” vs. “non-thinking” usage

• One flat rate regardless of input token size

Google said the adjustments reflect Flash’s strong performance and value relative to other models, aiming to simplify billing while maintaining competitive cost-efficiency.

Users still on the 2.5 Flash Preview 04-17 version can continue using it at its original rates until July 15, 2025, when the model will be retired.

Gemini 2.5 Pro Now Generally Available

Google also confirmed that Gemini 2.5 Pro, its most capable model, is now stable and generally available under the name gemini-2.5-pro.

The company reports that demand for Pro is the highest it has ever seen across any Gemini release. It continues to power advanced developer workflows, particularly in:

• Software development and coding

• Multi-step reasoning

• Agentic use cases (AI agents handling complex tasks)

The Pro model remains priced at the same rate introduced with the 06-05 preview. Developers using the earlier 2.5 Pro Preview 05-06 must transition by June 19, 2025, when that endpoint will be shut down.

What This Means

The Gemini 2.5 update signals more than just a product refresh—it reflects how foundation models are evolving into targeted, tiered tools designed for specific use cases. With Flash-Lite, Flash, and Pro now covering a spectrum of cost, speed, and reasoning depth, Google is aligning its model strategy with an industry-wide shift toward modular, scalable AI.

This mirrors a trend seen across leading AI companies: the move away from one-size-fits-all systems toward ecosystems of specialized models. Developers and enterprises increasingly expect options tailored for low-latency tasks, high-accuracy reasoning, or agentic workflows—and Google is responding by packaging intelligence into levels of performance that can be optimized for price or power.

Just as OpenAI and Anthropic have introduced models with selectable behaviors and capabilities, Gemini 2.5 brings that flexibility into a more fine-grained API-level control. The ability to adjust “thinking” on demand—controlling how much cognitive effort a model applies—gives users new levers to balance quality, speed, and cost in real-time applications.

As Gemini 2.5 continues to expand across use cases and budgets, Google is signaling its intent to lead not just on raw performance, but on flexibility, tooling, and deployment at scale.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.