Gemini 2.5: A sequence of thought mannequin updates

At the moment, we’re excited to share the most recent info on all the Gemini 2.5 mannequin household.

• Gemini 2.5 Professional is mostly accessible and secure (no modifications from 06-05 preview)

• Gemini 2.5 Flash is mostly accessible and secure (no modifications from 05-20 preview; see pricing replace under).

• Gemini 2.5 Flash-Lite now accessible in preview

The Gemini 2.5 mannequin is a pondering mannequin and might motive by thought earlier than responding, leading to higher efficiency and elevated accuracy. Every mannequin can management its thought funds, permitting builders to decide on when and the way a lot the mannequin “thinks” earlier than producing a response.

Introducing Gemini 2.5 Flash-Lite

At the moment we’re introducing a preview of two.5 Flash-Lite, which has the bottom latency and value of the two.5 mannequin household. It’s designed as a cheap improve from earlier 1.5 and a pair of.0 flash fashions. It additionally improves efficiency for many evaluations, lowering time to first token and rising tokens per decode per second. This mannequin is good for high-throughput duties corresponding to large-scale classification and summarization.

Gemini 2.5 Flash-Lite is an inference mannequin that lets you dynamically management your thought funds utilizing API parameters. Flash-Lite is optimized for price and velocity, so in contrast to different fashions, “pondering” is turned off by default. 2.5 Along with perform calls, Flash-Lite additionally helps all native instruments corresponding to Google search grounding, code execution, and URL context.

Gemini 2.5 Flash updates and pricing

Over the previous 12 months, our analysis group has continued to discover the Pareto frontier with our Flash mannequin sequence. When 2.5 Flash was first introduced, we had not but finalized the options of two.5 Flash-Lite. We additionally launched the product with a “doable value” and an “unthinkable value,” which brought about confusion amongst builders.

With the secure rollout of Gemini 2.5 Flash (this can be a preview of the identical 05-20 mannequin we made accessible at Google I/O) and the unbelievable efficiency of two.5 Flash, we’re updating the pricing for two.5 Flash.

• $0.30 / 1M enter tokens (*elevated from $0.15 enter)

• $2.50 / 1M output tokens (*lowered from $3.50 output)

• Eliminated the value distinction between pondering and non-thinking

• Maintained a single value vary no matter enter token dimension

Whereas we try to take care of constant pricing between preview and secure releases to reduce disruption, this can be a particular adjustment that displays the excellent worth of Flash and nonetheless offers one of the best price per intelligence accessible.

And with Gemini 2.5 Flash-Lite, we now have an excellent lower-cost choice (suppose it or not) for cost- and latency-sensitive use instances that require much less mannequin intelligence.

In case you are utilizing Gemini 2.5 Flash Preview 04-17, present preview pricing will stay in impact till deprecation scheduled for July 15, 2025, at which level mannequin endpoints might be turned off. You may transfer to the overall availability mannequin “gemini-2.5-flash” or swap to 2.5 Flash-Lite Preview for a decrease price choice.

Continued progress of Gemini 2.5 Professional

Progress and demand for the Gemini 2.5 Professional continues to be the quickest of any mannequin we have seen to this point. To allow extra prospects to construct this mannequin in manufacturing environments, we’re stabilizing the 06-05 model of the mannequin on the similar Pareto frontier value level as earlier than.

We predict Professional shines in instances the place you want probably the most intelligence and probably the most performance, corresponding to coding and agent duties. Gemini 2.5 Professional is on the coronary heart of lots of our most beloved developer instruments.

In case you are utilizing 2.5 Professional Preview 05-06, the mannequin might be accessible till June 19, 2025, after which will probably be turned off. In case you are utilizing 2.5 Professional Preview 06-05, merely replace the mannequin string to “gemini-2.5-pro”.

We will not wait to see extra domains profit from the intelligence of two.5 Professional and sit up for sharing extra about scaling past Professional within the close to future.