Google Unveils Enhanced Gemini 2.5 Pro: A Leap Forward in AI Capabilities

Google releases updated Gemini 2.5 Pro, says it's the "most intelligent model yet"

It seems like hardly a day goes by anymore without a new version of Google's Gemini AI landing, and sure enough, Google is rolling out a major update to its most powerful 2.5 Pro model. This release is aimed at fixing some problems that cropped up in an earlier Gemini Pro update, and the word is, this version will become a stable release that comes to the Gemini app for everyone to use. The previous Gemini 2.5 Pro release, known as the I/O Edition, or simply 05-06, was focused on coding upgrades. Google claims the new version is even better at generating code, with a new high score of 82.2 percent in the Aider Polyglot test. That beats the best from OpenAI, Anthropic, and DeepSeek by a comfortable margin. While the general-purpose Gemini 2.5 Flash has left preview, the Pro version is lagging behind. In fact, the last several updates have attracted some valid criticism of 2.5 Pro's performance outside of coding tasks since the big 03-25 update. Google's Logan Kilpatrick says the team has taken that feedback to heart and that the new model "closes [the] gap on 03-25 regressions." For example, users will supposedly see more creativity with better formatting of responses. Kilpatrick also notes that the 06-05 release now supports configurable thinking budgets for developers, and the team expects this build to become a "long term stable release." So, Gemini 2.5 Pro should finally drop its "Preview" disclaimer when this version rolls out to the consumer-facing app and web interface in the coming weeks.

Google says its updated Gemini 2.5 Pro AI model is better at coding | TechCrunch

Google on Thursday announced an update to its Gemini 2.5 Pro preview model that the company claims is better at certain programming tasks. The company's calling it an "updated preview," building on the upgrade to Gemini 2.5 Pro that Google announced around a month ago. Google says the model will roll out in general availability in a "couple of weeks," and is available starting today in its AI developer platforms AI Studio and Vertex AI and the Gemini app. "[Gemini 2.5 Pro] continues to excel at coding, leading on difficult coding benchmarks," Google writes in a blog post. "It also shows top-tier performance [on] highly challenging benchmarks that evaluate a model's math, science, knowledge, and reasoning capabilities." So what else is new? Google says it addressed feedback from its previous 2.5 Pro release, improving the model's style and structure. Now, 2.5 Pro can be "more creative with better-formatted responses," Google claims.

Google flexes AI muscle with Gemini 2.5 Pro updates - who doesn't love higher prompt limits?

The flagship model's rapid evolution reflects Google's growing confidence in its AI offerings. AI models improve in a positive feedback loop: Every user interaction provides more data to sharpen the model's output, leading to a better user experience and more frequent prompting, which in turn leads to better outputs, and so on. Gemini, Google's flagship generative AI model, has been riding this upward curve for quite a while. Now, some new updates to the model are signaling that Google is becoming a more confident competitor in the AI race. Also: The best AI image generators: Gemini, ChatGPT, Midjourney, and more Gemini App leader Josh Woodward announced in an X post Wednesday that paid subscribers to Gemini 2.5 Pro will now be able to submit up to 100 prompts to the model per day, up from the original limit of 50. Google introduced Gemini 2.5 Pro in March, describing it as the company's "most advanced model for complex tasks." Like OpenAI's o3 and DeepSeek's R1, Gemini 2.5 Pro is a "reasoning" model, so-called because it can analytically work through problems in a series of steps before providing its responses. Gemini 2.5 Pro was a major focus at Google I/O 2025, the company's annual developer conference last month. During the conference, Google announced that the model was being updated with "an experimental, enhanced reasoning mode for highly complex math and coding" called "Deep Think." Powerful models like Gemini and GPT-4 -- which powers ChatGPT -- require a huge amount of data and electricity to function. As a result, their developers will often limit the number of queries that users can ask per day to prevent computational overload. Imposing a daily query limit can also be a safety measure to ensure that new and experimental models don't veer into hallucinatory territory (which can be embarrassing and harmful to brands' reputations). Google's doubled daily query limit for Gemini 2.5 Pro signals that the company is confident in its new model's capabilities and that demand is continuing to rise: The Gemini App reportedly had more than 400 monthly active users as of late May; it still hasn't reached ChatGPT usage levels, but it's getting closer. One day before Woodward's X post, Google also elaborated on some audio features that had been initially unveiled a couple of weeks ago at I/O. Gemini 2.5 Flash preview -- an early iteration of a smaller, more nimble version of Gemini 2.5 -- is now able to engage in conversation that feels more natural and realistic, right down to subtle details like intonation and laughter. Users can access the 2.5 Flash preview through Google AI Studio, Vertex AI, or the Gemini App. Also: The pressure is on for quick AI rollouts - but slow and steady wins this race too Companies like Hume, which recently released the third iteration of its Empathic Voice Interface model, are also making advancements in this area. Users of Gemini 2.5 Pro and Flash preview now have much more creative freedom when generating text-to-speech (TTS) outputs. The models can be prompted, for example, to generate voices with specific emotions or expressive styles. "We believe conversation will be a key way we interact with AI," Google wrote in a company press release. Like virtually every other tech giant in Silicon Valley, Google was blindsided by the sudden release of ChatGPT in late 2022. At the time, it was widely regarded as a revolutionary moment for the industry. Since then, Google's been racing to catch up. But whereas a couple of years ago Google's generative AI efforts seemed like a thinly veiled effort to follow in the footsteps of OpenAI, its recent breakthroughs have given it a new air of dominance and confidence in the ongoing AI race. As its models continue to improve, so too will its competitive edge. Also: The hidden data crisis threatening your AI transformation plans (Disclosure: Ziff Davis, ZDNET's parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Google is easing up Gemini 2.5 Pro query limits for paid AI Pro users

During Google I/O, the tech giant's Gemini Advanced subscription plan, derived from the Google One AI Premium plan, was rebranded to Google AI Pro. The $20/mo plan unlocks Gemini in Workspace apps, Veo 3, additional storage, and even access to Gemini's latest and most capable plans, including 2.5 Pro and its Deep Research capabilities. Related Gemini 2.5 Pro, Google's 'most intelligent AI model,' is rolling out now But only as an Advanced perk Posts When first revealed back in March, Gemini 2.5 Pro was limited to users with a paid plan. The plan expanded to free users soon after, albeit with a smaller context window and rate limits in place. Now, in a bid to set the free and paid tiers further apart, Google is increasing Gemini 2.5 Pro's query limit for Google AI Pro users, letting heavy users utilize more of the AI model's upgraded reasoning, math, and coding capabilities. As pointed out by the folks over at 9to5Google, an apparent recent reduction in query limits led to a noticeable increase in users hitting the daily usage cap. The complaints have now prompted Google to essentially double the query limit for paying using. Source: Google For reference, Gemini 2.5 Pro's usage limits are currently described as 'limited access' for users without a paid plan, 'expanded access' for Google AI Pro users, and 'highest access' for those with a Google AI Ultra plan ($250/mo). Prompt away For users with a $20/mo AI Pro plan, Gemini 2.5 Pro's daily usage limit has been increased to 100 from 50 queries per day, as confirmed by Google VP Josh Woodward. "Thanks for using the model so much and messaging us wanting more," wrote Woodward in a tweet. Using the tech giant's most capable plan should now be less frustrating for users -- especially for those that actively pay for the plan instead of getting it for free as a perk with their new Pixel device.

The Gemini 2.5 Pro Preview inches closer to public release with a subtle, yet meaningful update

Summary Gemini 2.5 Pro just got another update, although it's still in Preview status and doesn't offer a paid tier with full access. Google cites multiple benchmark improvements and its attention to feedback as proof it's more effective than ever. Gemini 2.5 Pro and 2.5 Flash are slated for full public release sometime this month, but you can already access the Flash model in the Gemini app. Google unveiled the Gemini 2.5 Pro model in late March, followed by Gemini 2.5 Flash a month later. Both are still technically in Preview status rather than stable, and anybody can try out the 2.5 Flash model wherever the Gemini app is available. It's faster, more comprehensive, and more consistent than anything Google's put out yet. Gemini 2.5 Pro is a slightly different story. Only paid users can access it before it goes live, and the pricing is considerably higher than that of Google's other models. But Google is making every effort to ensure it's as capable as possible, with the recent update following hot on the heels of the upgrade it received in preparation for May's Google I/O event. Google touts some improved benchmark numbers to prove the current Preview model is the best in the business, but the real story could be the more subtle adjustments under the hood (Source: 9to5Google). Related Google Gemini can succeed where ChatGPT tasks fall flat Leverage the integration Posts 3 An awful lot of hype for a computer program This had better be good Google doesn't pull any punches regarding the purpose of the Gemini 2.5 Pro Preview. It's not aimed at the average consumer, there's no full-featured free tier, and the more lightweight Gemini 2.5 Flash is quite effective in its place. Google positions 2.5 Pro as its "state-of-the-art thinking model," capable of "analyzing large datasets, codebases, and documents using long context." It isn't meant for putting sunglasses on your dog, but getting serious work done with supposedly academic levels of rigor. 11:28 Related Can Google Search's cash cow live on in Gemini? [Video] Or will OpenAI and Perplexity eat that lunch? Posts To that end, Google's biggest claims around Gemini 2.5 Pro's latest update -- a "24-point Elo score jump on LMArena," a "35-point Elo jump to lead on WebDevArena at 1443," and "top-tier performance on GPQA and Humanity's Last Exam" -- won't mean much to most smartphone owners. But the model's intended market of developers and researchers will be at least somewhat familiar with those benchmarks. We also addressed feedback from our previous 2.5 Pro release, improving its style and structure -- it can be more creative with better-formatted responses. -- Google, on the latest Gemini update The real potential for improvement comes from the more intricate adjustments made, which Google tends to keep close to the vest. All Google really offered was the promise it addressed feedback, but that could mean a lot. Users have spoken up in recent weeks about decreased performance, due to whatever fine-tuning Deepmind was up to at the time. By taking feedback into account -- which is the point of a preview phase, after all -- Google should be able to improve Gemini performance in ways that benchmarks don't capture. With all that said, there's no way to know exactly what changes occurred behind the scenes with this upgrade. What we do know is that Google's most powerful, industry-focused model and its enhanced Thinking function, expansive Deep Research abilities, and field-leading benchmarks indicate it should continue to upend the professional world. And, considering Google has already relaxed query limits for paid Pro users, it might even be a little more affordable once it goes live later this month. Related 12 AI prompts I use to get more done in less time I used to spend hours on my tasks until I tried these prompts Posts 1

Google rolling out upgraded Gemini 2.5 Pro preview

Google today released an "upgraded preview" of Gemini 2.5 Pro that will soon be generally available in the coming weeks This model update builds off 2.5 Pro's I/O Edition (05-06) last month that brought coding upgrades. That coding lead continues with 06-05 today across benchmarks like AIDER Polyglot, while there's "top-tier performance on GPQA and Humanity's Last Exam (HLE), highly challenging benchmarks that evaluate a model's math, science, knowledge and reasoning capabilities." Meanwhile, Google notes how 06-05 today has a "24-point Elo score jump on LMArena" at 1470 today, while there's a "35-point Elo jump to lead on WebDevArena at 1443." With the last model upgrade, people noticed performance declines outside of coding. Google is now addressing that feedback with improved "style and structure," so that Gemini 2.5 Pro "can be more creative with better-formatted responses." This Gemini 2.5 Pro model update is still in preview, but you can expect generally available "in a couple of weeks." It is rolling out now to the Gemini API via Google AI Studio and Vertex AI, with developers able to control cost and latency with thinking budgets. The Gemini app is also getting this 2.5 Pro upgrade today.

Gemini app doubles recently introduced 2.5 Pro query limit

In recent days, Google AI Pro subscribers noticed lower query limits for 2.5 Pro in the Gemini app. That has now been doubled, with Google now specifying what the limits actually are. Historically, Google has not shared the specific number of daily prompts that paid users can send to Gemini's advanced models before they are capped. As of I/O 2025, when the rebranded AI Pro and new AI Ultra plans were introduced, the company just said users paying $19.99 per month get "Expanded access" (versus "Limited access" on the free tier) to 2.5 Pro (preview), while it's the "Highest access" when you pay $249.99/month. In general, most people -- save for the heaviest of users -- don't hit that restriction. Earlier this week, Google appeared to lower the limit, with many more heavy users -- especially coders that might go back-and-forth with Gemini -- hitting that wall. When that happens, they have to wait for the cap to reset, switch to 2.5 Flash, or upgrade to Google AI Ultra. Gemini chief Josh Woodward today acknowledged user complaints about the limit. There was no explicit acknowledgement about this week's change or why it was lowered (though Google continues to say that it sees "massive demand for Gemini 2.5"), but we learned that the just imposed limit was 50 queries per day. Google has now doubled the 2.5 Pro limit to 100 queries per day for the Gemini app in response to users "wanting more." 9to5Google's Take I don't think most casual users of the Gemini app will hit the limit unless they're working on a specific project that benefits from 2.5 Pro's "Reasoning, math & code" capabilities. Arguably, most queries can be addressed by the faster 2.5 Flash model, which became generally available in the Gemini app at I/O. Meanwhile, those that want the old experience likely have to subscribe to Google AI Ultra.

Google claims Gemini 2.5 Pro preview beats DeepSeek R1 and Grok 3 Beta in coding performance

Join the event trusted by enterprise leaders for nearly two decades. VB Transform brings together the people building real enterprise AI strategy. Learn more Google has released an updated preview of Gemini 2.5 Pro, its "most intelligent" model, first announced in March and upgraded in May, as a preview, intending to release the same model to general availability in a couple of weeks. Enterprises can test building new applications or replace earlier versions with an updated version of the "I/O edition" of Gemini 2.5 Pro that, according to a blog post by Google, is more creative in its responses and outperforms other models in coding and reasoning. During its annual I/O developer conference in May, Google announced that it updated Gemini 2.5 Pro to be better than its earlier iteration, which it quietly released. Google DeepMind CEO Demis Hassabis said the I/O edition is the company's best coding model yet. But this new preview, called Gemini 2.5 Pro Preview 06-05 Thinking, is even better than the I/O edition. The stable version Google plans to release publicly is "ready for enterprise-scale capabilities." The I/O edition, or gemini-2.5-pro-preview-05-06, was first made available to developers and enterprises in May through Google AI Studio and Vertex AI. Gemini 2.5 Pro Preview 06-05 Thinking can be accessed via the same platforms. Performance metrics This new version of Gemini 2.5 Pro performs even better than the first release. Google said the new version of Gemini 2.5 Pro improved by 24 points in LMArena and by 35 points in WebDevArena, where it currently tops the leaderboard. The company's benchmark tests showed that the model outscored competitors like OpenAI's o3, o3-mini, and o4-mini, Anthropic's Claude 4 Opus, Grok 3 Beta from xAI and DeepSeek R1. "We've also addressed feedback from our previous 2.5 Pro releases, improving its style and structure -- it can be more creative with better-formatted responses," Google said in the blog post. What enterprises can expect Google's continuous improvement of Gemini 2.5 Pro might be confusing for many, but Google previously framed these as a response to community feedback. Pricing for the new version is $1.25 per million tokens without caching for inputs and $10 for the output price. When the very first version of Gemini 2.5 Pro launched in March, VentureBeat's Matt Marshall called it "the smartest model you're not using." Since then, Google has integrated the model into many of its new applications and services, including "Deep Think," where Gemini considers multiple hypotheses before responding. The release of Gemini 2.5 Pro, and its two upgraded versions, revived Google's place in the large language model space after competitors like DeepSeek and OpenAI diverted the industry's attention to their reasoning models. In just a few hours of announcing the updated Gemini 2.5 Pro, developers have already begun playing around with it. While many found the update to live up to Google's promise of being faster, the jury is still out if this latest Gemini 2.5 Pro does actually perform better.

Google upgrades Gemini 2.5 Pro's already formidable coding abilities

The model is expected to become Gemini Pro's first official stable release Google's rapid rollout of new models of Gemini is continuing apace, but the latest version of Gemini 2.5 Pro has some notable improvements that the company claims will put it in play for a while as the first "long-term stable release." The upgrade also patches up some of the issues that might currently frustrate Gemini Pro users. For now, the model is still in beta, unlike its friskier sibling, Gemini 2.5 Flash. Gemini Pro 2.5 has reportedly taken longer and dealt with some complicated issues around regressions in conversation that made the AI model seem somewhat underpowered as a brain. Those issues have apparently been resolved, with Google bragging about the coding capabilities of the new model in particular. It's outscored rival models on the Aider Polyglot benchmark, a multi-language coding test popular for measuring AI model software composition. The updated model also offers developers what Google calls "configurable thinking budgets." These are essentially a way to fine-tune how much computing power Gemini uses to answer complex queries so that you don't use up all of your credits building an app in one go. It's already available through Google AI Studio and Vertex AI, and will likely become part of Gemini as a whole in the near future. And it's not just a technical whiz. Gemini Pro has faced complaints of lacking the same creative and conversational flair of its fellow Gemini models, failing to impress outside of technical tasks. The writing and formatting could get sloppy, and long-form responses tended to ramble or circle back on themselves. Google says it's fixed that issue, with correct formatting, more nuanced writing, and no trailing off mid-response. All of those upgrades lead to why Google has declared this version of Gemini Pro to be a long-term, stable model, at least for now. For developers and enterprise users, that kind of certainty is valuable in its own right, just as much as regular upgrades. The new model will have an impact on Gemini users outside of the office, too. The same improvements to formatting, memory, and contextual understanding will likely be incorporated into the public-facing version of Gemini just to keep things neat. And it fits with Google's strategy to embed Gemini everywhere and encourage everyone to use it for any of their AI needs. Gemini Flash is the default option for those not paying a subscription fee for Gemini. Gemini Nano handles AI for Android devices, but Gemini Pro is intended to be the flagship model, the one that impresses everyone. Google will definitely try to live up to that vision with the new model, but the competition has hardly gone away. OpenAI, Anthropic, and even Apple are all racing to be on top of the AI model game. Gemini 2.5 Pro proves Google won't be falling behind any time soon, at least now that it's stopped regressing.

Try the latest Gemini 2.5 Pro before general availability.

We're introducing an upgraded preview of Gemini 2.5 Pro, our most intelligent model yet. Building on the version we released in May and showed at I/O, this model will be the generally available, stable version starting in a couple of weeks, ready for enterprise-scale applications. The latest 2.5 Pro reflects a 24-point Elo score jump on LMArena, maintaining its lead on the leaderboard at 1470, and a 35-point Elo jump to lead on WebDevArena at 1443. It continues to excel at coding, leading on difficult coding benchmarks like Aider Polyglot. It also shows top-tier performance on GPQA and Humanity's Last Exam (HLE), highly challenging benchmarks that evaluate a model's math, science, knowledge and reasoning capabilities. We also addressed feedback from our previous 2.5 Pro release, improving its style and structure -- it can be more creative with better-formatted responses. Developers can start building with the upgraded preview of 2.5 Pro in the Gemini API via Google AI Studio and Vertex AI -- where we've also added thinking budgets to give developers more control over cost and latency. It's also rolling out today in the Gemini app.

Google Releases Upgraded Preview of Gemini 2.5 Pro, Better at Coding and Reasoning | AIM

The update builds on the version introduced in May and is intended for enterprise-scale use. Google has launched an upgraded preview of Gemini 2.5 Pro, which will become the stable, generally available version in the coming weeks. The update builds on the version introduced in May and presented at the Google I/O conference, and is intended for enterprise-scale use. The new model maintains its lead on multiple benchmarks. It shows a 24-point Elo increase on LMArena, taking its score to 1,470, and a 35-point jump on WebDevArena, now standing at 1,443. "The latest 2.5 Pro reflects a 24-point Elo score jump on LMArena...and a 35-point Elo jump to lead on WebDevArena," said Tulsee Doshi, senior director and product lead, on behalf of the Gemini team. Gemini 2.5 Pro also leads in coding benchmarks, including Aider Polyglot. It ranks high on academic benchmarks such as GPQA and Humanity's Last Exam, which test a model's capabilities in mathematics, science, knowledge and reasoning. The team addressed developer feedback on earlier iterations of 2.5 Pro, improving response formatting and output structure. "We also addressed feedback from our previous 2.5 Pro release, improving its style and structure -- it can be more creative with better-formatted responses," said Doshi. Developers can access the upgraded preview through the Gemini API via Google AI Studio and Vertex AI. Google has also introduced thinking budgets in these tools to help developers manage cost and latency. The model is also being rolled out in the Gemini app starting June 5. At Google I/O, the tech giant also introduced Deep Think for the Gemini 2.5 Pro models, further enhancing their reasoning capabilities. With Deep Think, the model scored 49.4% on the United States of America Mathematical Olympiad (USAMO), outperforming the Gemini 2.5 Pro without Deep Think (34.5%) and OpenAI's o3 (21.7%) and o4-mini (19.2%) models. Besides that Google also launched Gemini Diffusion, the company's new model that learns to generate outputs using methods used by image and video models, the company reported an impressive speed of 1,479 tokens/second.

Google revamps Gemini 2.5 Pro again, claiming superiority in coding and math - SiliconANGLE

Google revamps Gemini 2.5 Pro again, claiming superiority in coding and math Google LLC has launched another, even more capable preview of its powerful Gemini 2.5 Pro model, proclaiming it to be the "most intelligent" large language model it has released so far. It's the latest update to Gemini 2.5 Pro, which first debuted in March and was then upgraded as a preview in May, and it too is a "preview", with general availability slated to come in a couple of weeks. Google said companies can start building test applications with the new model, or alternatively replace earlier versions of Gemini 2.5 Pro in their existing applications. If they do, they may notice that it's more creative in its responses, and performs better when generating code and reasoning, the company claimed. Today's update builds on the revamped version of Gemini Pro 2.5 that was announced one month earlier at Google's I/O developer conference. That update was known as the "I/O edition", and today's version is called Gemini 2.5 Pro Preview 06-05 Thinking. According to Google DeepMind Chief Executive Demis Hassabis, it has achieved superior scores in coding benchmarks compared to the earlier models. For instance, the latest iteration has improved its score on the LMArena benchmark by 24 points, while its score in WebDevArena increased by 35 points, propelling it to the top of that ranking's leaderboard. In addition, Google released numerous benchmarks that show how it surpasses rivals such as OpenAI's o3, o3-mini and o4-mini, xAI Corp.'s Grok 3 Beta, DeepSeek Ltd.'s R1 and Anthropic's Claude 4 Opus in areas such as writing code, science and math. For instance, it demonstrated improved scores across AIDER Polyglot, GPQA, HLE and several other benchmarks. Google Chief Executive Sundar Pichai chimed in on X that the new model has also addressed some of the feedback and criticisms it received from its earlier Gemini 2.5 Pro releases, with improvements being made to its "style and structure" that enable it to be more "creative with better-formatted responses." Along with the frenetic pace of Gemini 2.5 Pro's update schedule, Google has been working to integrate the model into various applications and services. The latest version is available today in Google's "Deep Think" application, which carefully considers various hypotheses before responding, for example. Other ways to access it include the Gemini application programming interface, Google AI Studio and Vertex AI. It's priced at $1.25 per million tokens without caching for inputs, and $10 per million tokens for outputs, Google said. With the release of Gemini 2.5 Pro and the subsequent upgrades, Google has managed to steal back some of the limelight from rivals OpenAI and DeepSeek, which captured most of the headlines earlier this year with the first generation of their reasoning models.

Gemini 2.5 Pro AI Models Gets Upgraded Again With These Changes

The company said the model's responses were also being improved Google introduced an upgraded version of the Gemini 2.5 Pro artificial intelligence (AI) model on Thursday. Available in preview, the new model's technical name is "Gemini 2.5 Pro preview 06-05 Thinking" and it offers improved coding capabilities compared to the previous version that was released ahead of the Google I/O 2025. The Mountain View-based tech giant said this version also brings improvements in the style and structure of its responses. Notably, this is the second update the company has added to the model. In a blog post, Google announced the new version of the large language model, highlighting that it will be generally available in a couple of weeks. Notably, Gemini 2.5 Pro was originally released in March, which was then upgraded in May. The May upgrade was focused on the model's coding performance. The 06-05 version has also performed better on the LMArena, an AI model ranking platform where users rate models based on their experience of it. Google said the upgraded Gemini 2.5 Pro gets a 24-point Elo score jump compared to the previous version with a score of 1,470. It also increases its Elo score by 35 points on the WebDevArena and reaches 1,443. Google also highlighted the model's coding proficiency by mentioning that it is currently leading on the Aider Polyglot benchmark, considered to be a difficult coding-based test. The Gemini 2.5 Pro also keeps a similar score on the GPQA and Humanity's Last Exam (HLE) benchmarks, the company claimed. Apart from core capability boost, the tech giant is also said to have improved the style and structure of the model's responses. However, it is difficult to assess the improvements due to lack of any independent testers. While the stable version is still two weeks away, the upgraded preview version is now available in the Gemini app. Additionally, it can also be accessed with the Gemini application programming interface (API) via Google AI Studio and Vertex AI. Google said it has added thinking budgets to let developers have more control over the cost and latency.

Google Releases Upgraded Gemini 2.5 Pro in Preview | PYMNTS.com

By completing this form, you agree to receive marketing communications from PYMNTS and to the sharing of your information with our sponsor, if applicable, in accordance with our Privacy Policy and Terms and Conditions. This latest version is an enhanced version of the 2.5 Pro unveiled at its I/O developer conference in May, delivering even better performance when tested on popular benchmarks, Tulsee Doshi, senior director of product management at Google, wrote in a Thursday blog post. Doshi said the new model achieved a 24-point Elo score jump on LMArena -- this tests how well it does across a range of tasks -- putting it at the top of the leaderboard with a score of 1470. On WebDevArena, which evaluates web development skills, it rose by 35 points to lead at 1443. It also dominates the Aider Polyglot benchmark, which assesses complex coding capabilities across multiple languages. The model also ranks among the best on rigorous benchmarks like GPQA and Humanity's Last Exam -- both designed to test AI models on advanced reasoning, math, science and general knowledge. Google said developers can start building using the upgraded Gemini 2.5 Pro through Google AI Studio and Vertex AI. The enhanced AI model is also rolling out today in the Gemini app. Full general availability of Gemini 2.5 Pro is expected in coming weeks. Google said it made changes to the model in response to developer feedback. The release underscores Google's ongoing effort to lead in the rapidly moving AI landscape, particularly as OpenAI's ChatGPT is far ahead in usage and traffic. Google is doubling down on technical excellence, especially in areas like code generation and scientific reasoning. Last month, Google also gave developers a preview of Gemini 2.5 Pro, which it said has "significantly" improved capabilities for coding over the original release. Google said it released the model's preview earlier than expected due to "overwhelming enthusiasm." Gemini is Google's flagship family of AI models and the first major natively multimodal model, meaning it was built to be able to handle text, images, code and video from the ground up. Google first unveiled Gemini 2.5 in March as a "thinking" model capable of reasoning instead of just classifying and predicting data.

Google rolls out Gemini 2.5 Pro Preview for enterprise and developer use

On Thursday, Google announced the rollout of an updated Gemini 2.5 Pro model, building on the version introduced at the I/O 2025 event in May. The new Gemini 2.5 Pro shows a 24-point Elo score increase on LMArena, holding a top score of 1470 and maintaining its leaderboard position. It also gained 35 Elo points on WebDevArena, reaching 1443 and leading the rankings there. The model leads in coding benchmarks like Aider Polyglot and performs well on tests such as GPQA and Humanity's Last Exam (HLE), which measure math, science, reasoning, and knowledge. Tulsee Doshi, Senior Director of Product Management at Google, said the update reflects user feedback, with better creativity and improved response formatting. She added that "thinking budgets" have been introduced to help developers control cost and latency. Gemini 2.5 Pro ranks high across major AI benchmarks, outperforming OpenAI's o3 and o4-mini, Claude Opus 4, Grok 3 Beta, and DeepSeek R1. Key scores include: Gemini ranks among the highest performers across these categories. Its usage costs are also lower than many competitors. The rate is $1.25 per million input tokens, with caching adjustments ranging between $0.50 and $2.00, and $10.00 per million output tokens, with caching varying from $1.00 to $15.00. In comparison: Google plans to release Gemini 2.5 Pro as a stable, generally available model within a few weeks, aimed at enterprise-scale use. Meanwhile, developers can access the upgraded preview through Gemini API via Google AI Studio and Vertex AI. The Gemini app also started rolling out this version on June 5, 2025.