Google launches Gemini 3.5 Flash and 1 new video model to counter rivals

Google is escalating the AI arms race with a focus on speed and creative power, announcing at its I/O 2026 conference the Gemini 3.5 Flash model and a new video generator, Gemini Omni. The move comes as the company races to defend its territory against OpenAI and Anthropic, while new, highly efficient competitors emerge to challenge the industry’s cost structure.

"This represents a major leap forward in building more capable, intelligent agents," Google said of the new 3.5 model family in its announcement. The company claims Gemini 3.5 Flash achieves top-tier performance while maintaining the high speed needed for agentic workflows and serving teams of subagents, directly targeting the enterprise market’s need for scalable AI.

The announcements are part of a broad offensive to embed Gemini across Google's ecosystem, which now serves more than 900 million monthly users, up from 400 million the prior year. Alongside the new models, Google revealed a refreshed Gemini app and "Daily Brief," a proactive AI feature for subscribers that scans a user's inbox and calendar to organize their day.

At stake is leadership in a technology sector defined by a relentless pace of innovation and immense capital costs. While Google's new models aim to balance performance with speed, the emergence of hyper-efficient models from smaller players questions the long-term economics for enterprise customers, potentially shifting the basis of competition from pure capability to cost-per-query.

Gemini 3.5 Flash and Omni: Speed Meets Multimodality

Gemini 3.5 Flash is positioned as Google's fastest and most cost-effective model to date, designed to be the default across many of its services. The company claims it outperforms its predecessor, Gemini 3.1 Pro, on key coding and agentic benchmarks without compromising on intelligence. The model is available immediately in the Gemini app and through AI Mode in Google Search, with a more powerful Gemini 3.5 Pro version slated for release next month.

The more surprising announcement was Gemini Omni, a multimodal model that generates video from any combination of inputs, including text, images, audio, and other videos. This goes a step beyond Google's previous text-to-video model, Veo 3, by allowing users to edit and transform existing media through conversational prompts. Omni will be available for Google AI subscribers in the Gemini app and Google Flow, and will be offered for free to YouTube Shorts creators. To address safety concerns, Google is embedding SynthID digital watermarks in all Omni-generated videos.

A New Challenger Enters the Ring at 90% Lower Cost

Just as Google detailed its next generation of models, a two-year-old startup, Perceptron Inc., may have already redefined the efficiency frontier. The company launched its flagship video analysis model, Mk1, at an API price point that is 80-90% lower than current-generation flagship models from Google, OpenAI, and Anthropic.

Perceptron's Mk1 is priced at just $0.15 per million input tokens and $1.50 per million output tokens. According to company benchmarks, it matches or exceeds the performance of models like Gemini 3.1 Pro and GPT-5 on key video and spatial reasoning tasks. On the VSI-Bench for temporal reasoning, Mk1 scored 88.5, the highest among the models compared. This combination of high performance at a radically lower cost directly targets the industrial and enterprise markets for applications like robotics, security, and quality control.

The existence of a competitor like Perceptron highlights a critical challenge for incumbents like Google. While frontier models capture headlines, the battle for widespread enterprise adoption may be won by the provider who can deliver "good enough" intelligence at the lowest possible inference cost. For investors, the landscape is shifting from a pure performance race to a more complex equation where efficiency and accessibility are just as critical, impacting the valuation and revenue potential of the entire AI sector.

This article is for informational purposes only and does not constitute investment advice.