Trying to bypass competitors, Google presents a new, cheaper artificial intelligence model

Google office. Photo: tadviser.ru

As reported by the Financial Times, at its annual I/O conference in Mountain View on May 19, Google introduced the new Gemini 3.5 Flash model, with which the company hopes to close the gap on Anthropic and OpenAI in the artificial intelligence race. Company representatives stated that the new system is a significant step in the development of intelligent agents — programs capable of performing tasks independently based on user instructions.

Google claims that in terms of performance, Gemini 3.5 Flash is comparable to Claude Code models from Anthropic and Codex from OpenAI, but operates faster, and using the model could cost almost half as much. Next month, the company also plans to release a more scalable and powerful version, Gemini 3.5 Pro.

As FT notes, the launch of the new model comes amid a sharp increase in demand for computational power among AI laboratories. While some competitors face disruptions and are forced to limit client access to their services, Google is betting on its own chips. The company hopes this will allow it to maintain model operations with lower costs and greater stability than competitors.

Tools for code writing and business process automation are now considered the most promising way to monetize artificial intelligence. However, according to analysts' estimates, Google's share in this segment is only 10-15%, while smaller competitors control about 40% of the market.

The company Anthropic informed investors that it is nearing $45 billion in annual revenue, which is five times more than at the end of last year ($9 billion). This was almost entirely due to the popularity of its code-writing model and the Cowork product for business.

Google CEO Sundar Pichai was previously criticized repeatedly for allowing OpenAI and Anthropic to pull ahead, despite Google and DeepMind's years of leadership in AI research.

Pichai is now betting on expanding the use of intelligent agents. While such tools were previously intended primarily for developers and corporate clients, the company now wants to make them accessible to mass users.

This strategy relies on a so-called "full cycle" — in-house chip production, data center construction, creation of foundational models, and development of end products. Google plans to spend up to $190 billion this year on the development of this infrastructure.

Prospects for Google Search

As reported by FT, special attention at the conference was given to restructuring the Google search engine. The company continues to transform search into a platform for working with AI. Users will now be able to create intelligent agents directly in the browser. Such bots can independently search for news, select products, or book tickets.

The company intends to embed programming tools into search, allowing users to create their own interfaces and model various scenarios.

Previous innovations — "AI Mode" and short "AI Overviews" — are already used by over a billion people monthly. However, this has led to a decrease in traffic to original source websites, causing dissatisfaction among publishers and advertisers.

Despite the number of users of the Gemini application growing to 900 million per month over the year, the service still lags behind ChatGPT in audience size, which has the same number of users in a week.

To intensify competition, Google is reducing the price of its premium AI subscription to $200 per month. This matches OpenAI and Anthropic's rates, but Google offers higher usage limits and additional services, including YouTube and several terabytes of cloud storage.

Thus, the company is trying to use its scale and financial capabilities to increase pressure on smaller competitors, who are still operating at significant losses.

Among other announced products, Gemini Omni stands out — a multimodal model capable of creating and editing video based on text, images, and voice commands.

• 3
• 1
• 0
• 0
• 0
• 0
• 0
• 0