Openai: Claude OPUS 4, seven hours without stopping, and records Swe-Bench and Reshapes Enterprise Ai

man Absolute Claude Obus 4 and Claude Sony 4 Today, the tape is greatly raised by what artificial intelligence can accomplish without human intervention.

The pioneer of the company OPUS 4 Form Focus maintaining a complex open source rebuilding project for about seven hours during the test in Two Hypocrisy transforms artificial intelligence from a quick response tool to a real collaboration capable of addressing projects that last day.

The performance of a jump marathon is an amount that exceeds the attention of the minutes that extend over the previous artificial intelligence models. Deep technological effects: Artificial intelligence systems can now be dealing with complex software engineering projects from pregnancy to end, maintaining context and focusing throughout the entire day of work.

Antarbur claims Claude Obus 4 Achieve 72.5 % on Beach seatsStandard of strict software engineering, outperform Openai’s GPT-4.1Which recorded 54.6 % when it was launched in April. The achievement proves Anthropor as a huge competitor in the increasingly crowded artificial intelligence market.

Beyond quick answers: The Revolution of logic turns Amnesty International

The artificial intelligence industry has greatly achieved thinking models in 2025. These systems operate through problems systematically before responding, and to simulate human -like thoughts instead of just matching patterns against training data.

Openai began this shift with “O” series Last December, followed by Google’s Gemini 2.5 Pro With experimental “Deep thought“The ability. Depsik R1 Model The market share has been unexpectedly captured with its exceptional capabilities to solve problems at a competitive price point.

This axis indicates a fundamental development in how people use Amnesty International. According to Bo Spring 2025 directions to use the artificial intelligence model The report, the use of the thinking model jumped five times in just four months, grows from 2 % to 10 % of all artificial intelligence reactions. Users are increasingly viewing artificial intelligence as an intellectual partner of complex problems instead of a simple questions system.

The new Claude models distinguish themselves through integration Use the tool Directly in the thinking process. The simultaneous research and seasonal approach reflects this human awareness more closely than the previous systems that collected information before starting the analysis. The ability to stop and search for data and integrate new results during the thinking process creates a more natural and effective experience to solve problems.

Architectural engineering balances with dual mode with depth

Antarbur took a continuous contact point in the experience of the artificial intelligence user with it Hybrid approach. Both Clauds 4 offer semi-fixed responses to direct information and expanded thinking of complex problems-eliminating the improved delays in previous thinking models imposed on simple questions.

This double mode function maintains Snappy reactions that users expect while opening deeper analytical capabilities when needed. The system is dynamically devoted to thinking resources based on the complexity of the task, as it achieved a balance that has failed to achieve previous thinking models.

Memory stability It stands as another penetration. CLADE 4 models can extract the main information from documents, create brief files, and maintain this knowledge through sessions when giving appropriate permissions. This capacity is a “problem of memory loss” that limits the benefit of artificial intelligence in long -term projects where the context must be maintained over days or weeks.

Technical implementation is similar to how human experts are developing on knowledge management systems, as artificial intelligence automatically organizes information in improved organization formats to recover the future. This CLADE approach allows to build an increasing understanding of the increasingly complex areas of the extended reaction periods.

The competitive scene with the battle of artificial intelligence leaders for the market share

The timing of a person’s declaration highlights the frequency of rapid competition in advanced artificial intelligence. Just five weeks after Openai GPT-4.1 familyThe anthropoor faced the models that it challenged or exceeded in the main standards. Google updated it Gemini 2.5 assortment Earlier this month, while Meta recently released Lama 4 models It includes multimedia capabilities and a symbolic context of 10 million.

Each major laboratory has been carved distinguished strengths in this increasingly specialized market. Openai performs in General thinking and The integration of the toolGoogle excels in UnderstandingIt is now demanding that the crown for continuous and professional performance Coding applications.

The strategic effects of institutions are great. Institutions are now facing increasingly complicated decisions on artificial intelligence systems to publish specified use cases, with no one model that dominates all measures. This retail benefits from developed customers who can take advantage of specialized strengths of artificial intelligence with the challenge of companies looking for simple and united solutions.

The anthropier expanded Claude icon. The system now supports background tasks across Gaybab’s verbs And it is original with Opposite the code and Jetbrains Environments, display the adjustment of the proposed software instructions directly in the files of the developers.

Gaytap’s decision to integrate Claude Sonnet 4 as a key model for a new coding agent in Copilot It provides the market health. This partnership with the Microsoft Development platform indicates that large technology companies are diversifying artificial intelligence partnerships instead of relying exclusively on individual service providers.

Antarbur has completed its typical versions with new API capabilities for developers: the enforcement instrument tool, the MCP connector, file programming interface, and rapid storage for up to an hour. These features allow the creation of the most advanced artificial intelligence agents that can continue through the complex workflow – essential for the adoption of institutions.

Transparency challenges appear as models grow more sophisticated

Aprili search paper in April, “Thinking models do not always say what they think“The patterns of how these systems are revealed to their thinking. Their studies were found Claude 3.7 Sonata Decisive hints mentioned that they were used to solve problems only 25 % of the time – which raises important questions about the transparency of the logic of artificial intelligence.

This research pursues an increasing challenge: when models become more capable, they also become more mysterious. The seven -hour independent coding session, which offers endurance in Claude Obus 4, also shows how difficult people check these extensive logical chains.

The industry is now facing a paradox as it brings increased ability to decrease transparency. Treating this tension will require new curricula to supervise artificial intelligence that balances performance with the ability to clarify – the anthropier challenge itself recognized but has not been completely solved yet.

The future of the continuous artificial intelligence cooperation is formed

Independent working session between Claude Obus 4 seven hours glimpse of AI’s future role in knowledge works. With the development of extended focus models and improved memory, they are increasingly similar to collaborators instead of tools – capable of continuous complex work with the minimum human supervision.

This progress indicates a profound shift in how institutions create knowledge work. Tasks that require continuous human interest can now be delegated to artificial intelligence systems that maintain focus and context for hours or even days. The economic and regulatory effects will be significant, especially in areas such as software development, as talent deficiency continues and employment costs remain high.

Since Claude 4 blur the line between human and automatic intelligence, we are facing a new reality in the workplace. The challenge that we face is no longer wondering whether artificial intelligence can match human skills, but adaptation to a future as our most productive colleagues may be digital and not a human being.

Daily visions about business use cases with VB daily

If you want to persuade your boss at work, you have covered VB Daily. We give you the internal journalistic precedence over what companies do with obstetric artificial intelligence, from organizational transformations to practical publishing operations, so that you can share visions of the maximum return on investment.

Read with us privacy policy

Thanks for subscribing. Check more VB news bulletins here.

An error occurred.