Explore improvements in the stable and generally available releases of Gemini 2.5 Pro and 2.5 Flash, along with a new preview for 2.5 Flash-Lite, now available in the Gemini API. Gemini 2.5 models bring configurable thinking budgets to give you more control over performance and cost.
Introducing 2.5 Flash-Lite, our new cost-efficient reasoning model optimized for high-throughput, low-latency tasks. Build faster, more powerful agentic applications with native support for tools such as Grounding with Google Search, URL context, and code execution.
Balancing speed and performance, this workhorse model handles multimodal, coding, and long-context tasks at scale. To simplify planning and reflect the cost efficiency of 2.5 Flash-Lite, we’re updating 2.5 Flash pricing to $0.30 / 1M input tokens and $2.50 / 1M output tokens, removing the thinking versus non-thinking price difference.
The 2.5 Flash Preview 04-17 model will no longer be supported after July 15, 2025. You can transition to the generally available model or switch to 2.5 Flash-Lite Preview as a lower-cost option.
Now production-ready, our most intelligent model tops the LMArena and WebDev Arena leaderboards and receives even higher marks on Aider Polyglot, GPQA, and HLE than the previous version. We’ve also added thinking budgets for more control over how much the model reasons.
We’re excited for you to integrate the latest Gemini 2.5 models into your applications! Read the blog post to learn more.
Happy building,
The Google AI Studio team
Building Coding Capabilities
Get the inside scoop on how the Gemini team built one of the world’s best coding models and what’s next in the latest episode of Google AI Release Notes.
This email was sent to va@taara.in because you signed up to receive emails about Google AI. If you do not wish to receive these emails, please unsubscribe.
My learnings from Davos
I just got back from Davos. For all its polish, the conversations were very real.
A Message from Andrew Feldman, Founder and CEO
Greetings friends and supporters. It's Andrew.
From now on, I'm going to send you the occasional email when I have something worth sharing.
I just got back from Davos. For all its polish, the conversations were very real.
In side rooms and late dinners, global leaders kept c...
Viva Las Cerebras! A special offer for AWS re:Invent
Join us for exclusive offers and events
Headed to AWS re:Invent in Las Vegas, December 1–5? There are no sure bets in Vegas, but the odds are stacked in your favor when you build on the world’s fastest AI infrastructure (20x faster than NVIDIA GPUs) through AWS Marketplace!
Here’s what you need to know:
The Offer: Don't step foot on the expo floor until you've tried the fastest frontier open models like GLM-4.6, OpenAI's GPT OSS 120B & more, powered by Cerebras Inference on AWS Marketplace — no new vendor approvals, no procurement, just use your existing account and pay-as-you-go. While s...
Comments
Post a Comment