It was several months late, but Google finally unveiled Gemini 4

Google has announced a new top-level model called Argon. This is in the Gemini 4 series and is designed to handle the most complex of tasks. The company spokesperson said the size of this model is larger than the last Pro series. Larger models need more computing power. Google hopes to gain stronger understanding capabilities by increasing the scale. The more difficult the task, the more differences can be observed between different models.

The release date of this model was several months later than expected by the outside world. The originally planned Gemini 3.5 Pro is no longer being released. Google has not given a specific time for the full release of Argon. Enterprise users are still waiting for a clear schedule. During this waiting period, many teams have already tried out the products of competitors.

Google made an adjustment to the DeepMind laboratory recently. It was disclosed that the founder and CEO, Demis Hassabis, resigned, and several Gemini model leaders also left. The development of large models relies heavily on team synergy. After key personnel leave, many processes need to be re-adjusted. At the same time, OpenAI and Anthropic released new versions. For every delay by Google, the competitors have one more chance to seize the market.

Argon's focus is not on daily chatting, but on coding and detecting vulnerabilities. Google claims that it can match OpenAI's Astra and Anthropic's Opus in key programming and security evaluations. However, the figures released by Google are not fully leading. In four programming-related evaluations, Argon is behind in two aspects. There are also differences in some other indicators. The performance in real projects should be judged by programmers and security engineers.

CEO Sundar Pichai had refuted the claim that Google was lagging behind in July. But in recent months, the focus of promotion has changed. In the past, it was about how strong the capabilities of Gemini were. Now, it's about how low the usage cost is. This shift is very realistic. Enterprises do not only consider the effectiveness when purchasing AI, they also consider who can complete a task more efficiently and who can be called more frequently on a large scale.

Larger models mean higher operating costs. As the number of calls increases, the bill will rise rapidly. Google has regarded price as a selling point and has begun to attach importance to the actual implementation effect. But being cheap can not be achieved by sacrificing quality. If the model frequently makes errors, the money spent on manual corrections may be more than the savings from reduced call fees. This is something enterprises are well aware of.

Argon will not be made available to everyone immediately. Google first hands it over to selected cybersecurity partners. These institutions can use the model to check for system vulnerabilities and assess defense capabilities. The powerful AI has both positive and negative aspects in the field of security. It can help people strengthen their defenses, but it can also be used to identify weaknesses. Before expanding the openness, manufacturers need to repeatedly assess these risks.

Google also took part in the voluntary pre-release access process that was led by the Trump administration. That gives regulators a chance to get a head start on understanding what the model can do. It is not a mandatory approval process, more like a communication mechanism. Enterprises can identify problems in advance, and the government can also keep track of technological progress. For ordinary users, there will be an observation period before the official version is launched.

Training a top-level model requires a large number of chips and electricity. The larger the model, the higher the cost of training once. Manufacturers need to purchase equipment, build data centers and maintain teams. This investment is not a one-time expense. Competitors update monthly, and one has to keep up with the pace. Teams with insufficient funds are difficult to remain on the table for the long term. Google's scale advantage is evident here. The competition for talent is equally fierce. The flow of top researchers often determines the direction of the next generation of technologies. Google's recent personnel changes have raised concerns from the outside world.

AI competition is shifting from comparing parameters to practicality. Google has search, cloud services and the Android ecosystem. These channels enable models to quickly reach a large number of users. But channel advantages can not replace technical strength. Developers will vote with their feet, using whichever tool is convenient. The real results of Argon will only be known when users provide answers. Release is just the first step. Google still needs to use product experience to make up for the several months of delay.