Google Announces Gemini 3.5 Model, Enhanced for AI Agents
At its I/O annual developer conference on May 19, Google unveiled the new-generation AI model series, Gemini 3.5, aiming to compete with OpenAI and Anthropic in both consumer and enterprise markets. The series is optimized for executing "agent tasks," with the first model, 3.5 Flash, now available to global users, offering higher speed and cost-effectiveness. Concurrently, Google introduced the multimodal model Gemini Omni for video generation and editing, emphasizing content transparency through digital watermarking technology.
📋 Article Processing Timeline
- 📰 Published: May 20, 2026 at 02:53
- 🔍 Collected: May 20, 2026 at 03:01 (8 min after Published)
- 🤖 AI Analyzed: May 20, 2026 at 03:05 (3 min after Collected)
The Google I/O annual developer conference kicked off today, featuring the announcement of the new-generation AI model series, Gemini 3.5, which will challenge OpenAI and Anthropic in both the consumer and enterprise application markets. Google emphasized that this series is specially optimized for executing "agent tasks." The first model, 3.5 Flash, is available to users worldwide starting today. This means that when general users use the Gemini App and Google Search's AI mode, 3.5 Flash will serve as the default underlying model.
Google also announced Gemini Omni, a multimodal video generation and editing model that can be considered a video version of its image model, Nano Banana.
● 3.5 Flash Available Today, 3.5 Pro Expected in June
Regarding the core direction of this year's I/O, Google CEO Sundar Pichai stressed that the development of artificial intelligence (AI) has entered a phase of rapid advancement. Users now place more importance on the practical value of AI within products rather than just technical demonstrations.
The new 3.5 Flash model combines "frontier intelligence" with "action capability," the latter being a key core of "AI agents."
In addition to general users, developers can access it through Google's Antigravity development platform, as well as the Gemini API in Google AI Studio and Android Studio. Furthermore, 3.5 Pro is already in use internally at Google and is expected to be launched externally in June.
3.5 Flash boasts advantages in speed and cost. According to Google, based on the number of tokens output per second, it is about 4 times faster than other frontier models, at less than half the cost, and in some cases, the cost can be reduced to about one-third.
Additionally, in almost all types of benchmarks, the performance of 3.5 Flash surpasses the previous flagship model, Gemini 3.1 Pro.
● Omni Multimodal Generation Model Turns Ideas into Cinematic Videos
Besides the main model, Google also unveiled the Gemini Omni multimodal generation model at this year's I/O, noting that Omni can be thought of as an image model like Nano Banana, but primarily for creating videos.
At present, Google is focusing on video generation and editing. Ultimately, the hope is to develop Omni into a world model capable of receiving images, sounds, videos, and text, and using its understanding and reasoning abilities to produce various output formats, including video.
The first model in the Gemini Omni series, Omni Flash, is available starting today to Google AI Plus, Pro, and Ultra subscribers worldwide through the Gemini App and Google Flow. It will also be available for free to users of YouTube Shorts and the YouTube Create App starting this week. The next advanced version, Omni Pro, is expected to be released soon.
Google also emphasized that content generated or edited through Omni will automatically include a SynthID digital watermark. This allows users to clearly identify which content is AI-generated or edited with AI tools, thereby enhancing the overall transparency and security of the content.
Google also announced Gemini Omni, a multimodal video generation and editing model that can be considered a video version of its image model, Nano Banana.
● 3.5 Flash Available Today, 3.5 Pro Expected in June
Regarding the core direction of this year's I/O, Google CEO Sundar Pichai stressed that the development of artificial intelligence (AI) has entered a phase of rapid advancement. Users now place more importance on the practical value of AI within products rather than just technical demonstrations.
The new 3.5 Flash model combines "frontier intelligence" with "action capability," the latter being a key core of "AI agents."
In addition to general users, developers can access it through Google's Antigravity development platform, as well as the Gemini API in Google AI Studio and Android Studio. Furthermore, 3.5 Pro is already in use internally at Google and is expected to be launched externally in June.
3.5 Flash boasts advantages in speed and cost. According to Google, based on the number of tokens output per second, it is about 4 times faster than other frontier models, at less than half the cost, and in some cases, the cost can be reduced to about one-third.
Additionally, in almost all types of benchmarks, the performance of 3.5 Flash surpasses the previous flagship model, Gemini 3.1 Pro.
● Omni Multimodal Generation Model Turns Ideas into Cinematic Videos
Besides the main model, Google also unveiled the Gemini Omni multimodal generation model at this year's I/O, noting that Omni can be thought of as an image model like Nano Banana, but primarily for creating videos.
At present, Google is focusing on video generation and editing. Ultimately, the hope is to develop Omni into a world model capable of receiving images, sounds, videos, and text, and using its understanding and reasoning abilities to produce various output formats, including video.
The first model in the Gemini Omni series, Omni Flash, is available starting today to Google AI Plus, Pro, and Ultra subscribers worldwide through the Gemini App and Google Flow. It will also be available for free to users of YouTube Shorts and the YouTube Create App starting this week. The next advanced version, Omni Pro, is expected to be released soon.
Google also emphasized that content generated or edited through Omni will automatically include a SynthID digital watermark. This allows users to clearly identify which content is AI-generated or edited with AI tools, thereby enhancing the overall transparency and security of the content.