Crypto news

While OpenAI and Anthropic compete to create AI capable of recursive self-improvement, Google DeepMind is making a different bet — on "World Models" that understand the physical laws of reality. Outwardly, the difference is barely noticeable: Google continues to release models and generate revenue. But if you look at the positions in independent rankings, it becomes obvious: the tech giant is deliberately leaving the race for "raw" power.

Gemini 3.6 Flash: Betting on Speed and Price, Not Leadership

On July 21, the Gemini 3.6 Flash model was released. The main focuses are speed and cost, not maximum performance. According to Google, the new version generates 17% fewer tokens than its predecessor, which directly reduces the cost per response. However, in the Artificial Analysis ranking, this model only took 10th place. All leading developers ranked higher.

Google has not abandoned refinements. The company has launched the largest training run for the future Gemini 4, and a larger-scale version, Gemini 3.5 Pro, is currently undergoing testing with partners. But CEO Sundar Pichai has already linked the roadmap to the development of personal agents, not to positions in rankings. Investors reacted more sharply: in June, Alphabet's shares fell by 6% after two leading researchers left for competitors.

What DeepMind Risks

A world model is an AI that studies the physical patterns of the environment: gravity, motion, cause-and-effect relationships. The algorithm predicts the development of events in space, rather than simply selecting words. DeepMind openly declares this course: on the company's website, both Genie 3 and Gemini Robotics are located in the world models and embodied AI section.

In May, the lab expanded Project Genie, integrating it with Street View, and introduced SIMA 2 — an agent that learns through interaction with virtual 3D worlds.

OpenAI and Anthropic have chosen a completely different strategy. These teams are focused on recursive self-improvement of systems. Essentially, this technology allows AI to independently design the next generation of algorithms. The product launches a new development cycle, creating even more powerful solutions.

Analyst Alberto Romero suggested that Google consciously decided not to participate in this race: "Hassabis is betting on something else — world models capable of understanding and simulating the real world, not just predicting the next token." However, Google has not made such statements.

Why a Competitor's Co-Founder Considers DeepMind an "Exception"

The most critical view was voiced by a representative of competitors several months ago. Anthropic co-founder Jack Clark published an essay on the future of AI on May 4. He estimated the probability that AI could independently manage the research process by the end of 2028 at 60%. For 2027 — 30%.

Clark then questioned which labs are truly moving towards this goal. According to him, DeepMind "looks the most restrained among the big three." Restraint implies caution. Clark's arguments are based on DeepMind's own documents, particularly an article on AI safety issues published in 2025.

Anthropic acts more boldly. Research on recursive self-improvement states that by May 2026, Claude wrote over 80% of the code that the company deploys in its products. Before February 2025, this figure was close to zero. The company has already recorded cases where AI creates more advanced AI: over one testing period, their models increased performance 52-fold in April. A year earlier, this increase was only 2.9 times. For comparison, an experienced engineer would need four to eight hours for a fourfold increase.

Why It Can't Be Said That Google Has Completely Left the AI Race

There are two important factors contradicting this hypothesis:

• Google holds the lead in machine learning research tests.
• The system demonstrates high user activity in the global market.

The MLE-Bench benchmark evaluates models' ability to independently create AI architectures. In February, Gemini 3 achieved a score of 64.4%, taking first place. Later, in July, version 3.6 Flash scored 63.9%. Companies exiting the market usually do not top such rankings.

Furthermore, the tech giant maintains enormous reach. Sundar Pichai stated that the Gemini app audience reached 950 million monthly users.

Can Google Afford Delays?

Revenue from search advertising has long covered experimental expenses, allowing DeepMind to work without undue haste. However, recent Alphabet reports indicate a shrinking financial cushion.

In the second quarter of 2024, the company's revenue reached $119.8 billion, 24% higher than the previous year. According to the July 22 report, search advertising alone brought in $63.3 billion. However, expenses are rising. Over three months, Alphabet allocated $44.9 billion to developing data centers and purchasing equipment — almost double the amount from a year ago.

As a result, free cash flow was negative $5.86 billion. In March, this figure was positive at $10.1 billion. In December of the previous year, it was $24.6 billion. To cover the difference, Alphabet issued $49.6 billion in shares and raised $20.3 billion in loans. Long-term liabilities grew from $46.5 billion to $98.2 billion over six months.

The expense item related to joint AI developments resulted in a loss of $5.79 billion — a year ago it was $3.37 billion. Consequently, the current wait-and-see position is costing the corporation increasingly more.

What to Watch in the Next 30 Days

• Whether the release of Gemini 3.5 Pro will take place and what positions the model achieves in tests.
• Whether DeepMind will present new data on world models for Gemini 4.
• Whether Alphabet's cash flow will return to positive values in September.
• Whether Demis Hassabis will articulate a clear position regarding self-learning technology.

The launch of Gemini 4 will be a defining moment for the company's strategy. If the approach to creating world models proves itself, the current slowdown will turn out to be a precise calculation. The upcoming financial report will show the real limits of Alphabet's resources in this technological race.

Expert Opinion: Google is not losing — it is changing the rules of the game. While competitors invest billions in recursive self-improvement, DeepMind is making a long-term bet on understanding the physical world. If world models truly become the next breakthrough, Google's current "restraint" could turn into a strategic advantage. However, the market and investors demand results here and now — and Alphabet will have to prove that its path is not a dead end.