Anthropic’s new Claude 4.1 dominates coding tests days before GPT-5 arrives

Anthropor released a developed version of the pioneering artificial intelligence model on Monday, as it achieved new performance in software engineering tasks where artificial intelligence start races to maintain their hegemony in the profitable coding market before the expected competitive challenge from Openai.

The new CLADE OPUS 4.1 recorded 74.5 % on Swe-Becked, a wide-scale monitored standard that tests the ability of artificial intelligence systems to solve software engineering problems in the real world. The performance of the Openai O3 exceeds 69.1 % and Guemini 2.5 Pro from Google by 67.2 %, which enhances a pioneering position in the anthropologist in the Acting Coding Aid.

The version comes at a time when Anthropor has achieved amazing growth, as the annual repeated revenues have achieved five times from one billion dollars to 5 billion dollars in only seven months, according to industry data. However, the company’s height in the company has created a dangerous dependency: approximately half a $ 3.1 billion of API revenue stems from only two customers – an assistant coding index and GitHub Copilot from Microsoft – generating $ 1.4 billion combined.

“This is a very frightening situation in which you should be. Changing one decade and you are subject,” Guillaume Leverrier, the first product manager at Logitech, in response to social media concentration data.

Openai and Anthropic appears amazing growth in 2025, where Openai Arr has doubled in the past six months from 6 billion dollars to $ 12 billion and an anthropologist increased 5x from one billion dollars to $ 5 billion in 7 months.

If we compare revenue sources, the image is very interesting:
– Openai … pic.twitter.com/8oan1rsm9e

Peter Gusif (Petergostev) August 4, 2025

The upgrade represents the last step from the anthropoor to fortify its location before it launches Openai GPT-5, is expected to challenge the superiority of Claude coding. Some industrial monitors asked whether the timing indicates urgency rather than preparing.

“OPUS 4.1 appears to be like the accelerated version of GPT-5”, and compare the model inadvertently with competitors at the user interface tasks, “wrote Alec Felicanov. The comment reflects broader speculation in the industry that the anthropoor accelerates its release schedule to maintain the market share.

How two clients generate nearly half of API revenues of $ 3.1 billion

The Anthropier’s human business model has become increasingly focused on software development applications. The company’s Claud Code subscription service, at a price of $ 200 per month compared to $ 20 for consumer plans, reached $ 400 million of repeated annual revenues after doubling in only weeks, indicating the appetite of the enormous institutions of artificial intelligence coding tools.

“Claude Code, which makes 400 million in 5 months, with no marketing spending is a type, right?” Minh Nhat Noguyen, with highlighting the rate of organic adoption between professional programmers.

Well, Claude Code, which makes 400 million in 5 months with no marketing spending mainly is somewhat crazy, right? https://t.co/hiy34qdluq

– Minh Nhat Nguyen (@Menhguin) Aug 5 2025

The concentration of profitable coding has proven to be risky. While Openai dominates the revenues of subscriptions in consumers and businesses with broader applications, Antarbur may intenrate a leadership position in the developer market. Industry analysis shows that “every one coding assistant is behind Claude 4 Sonnet,” according to Peter Gosev, who tracks the revenues of the IQ.

GitHub, which Microsoft acquired for 7.5 billion dollars in 2018, represented a particularly complicated relationship with Anthropor. Microsoft has a large share in Openai, which creates potential conflicts as GitHub Copilot depends greatly on human embeddles while Microsoft has the competing Amnesty International capabilities.

“I am not, one of these is 49 % owned by a competitor … so there is for weakness as well,” Sia Mali, a business colleague, noticed, referring to the Microsoft ownership structure.

Code -reinforced Code’s capabilities with more strict safety protocols come after artificial extortion tests

Besides coding improvements, OPUS 4.1 enhances research capabilities and data analysis on Cloud, especially in detailing independent search functions. The model maintains a hybrid thinking approach in humans, as it combines direct treatment with expanded thinking capabilities that can use up to 64,000 symbols of complex problems.

However, the progress of the model comes with the increasing safety protocols. OPUS OPUS 4.1 under the AI’s 3 (ASL-3) safety framework, the most applied appointment by the company has applied, which requires reinforced protection against the theft of models and abuse.

Previous tests of CLADE 4 models were revealed in relation to behaviors, including extortion attempts when artificial intelligence believed that she had faced the closure. In the scenarios subject to control, the model threatened to reveal personal information about engineers to maintain its existence, indicating advanced but dangerous thinking capabilities.

Safety concerns were not deterring institutions ’adoption. GitHub reports that CLAUDE OPUS 4.1 offers “particularly noticeable performance gains in reshaping a multi -file code”, while the Rakuten collection accurately praised the “determining micro -corrections within the large code bases without making unnecessary modifications or making errors.”

Why the GPT-5 of Openai is an existential threat to the Anthropor strategy that focuses on the developer

The artificial intelligence coding market has become a high -risk battlefield of billions of dollars. The developers’ productivity tools are some of the clearest instant applications for the Troy AI, with measurable productivity gains that justify the outstanding pricing of institutional customers.

Anthropor’s concentrated customer base, despite profitable, creates weakness if it is possible for competitors to attract the main customers. The coding assistant market is especially preferred to switch fast models, as developers can easily test the new AI systems through simple API changes.

“My feeling is that the growth of anthropologist relies heavily on its dominance in coding,” Gustyv noted. “If GPT-5 challenges it, with EG Cursor and Github Copilot to Openai, we may see some reflection on the market.”

Competitive dynamics may intensify with decreased costs of devices and improvement of reasoning improvements, which may potentially be capabilities of artificial intelligence over time. “Even if there is no improvement for coding a coding model from all artificial intelligence laboratories, the low HW costs and the improvement of Inf improvements alone will lead to profits in about 5 years,” Fenkat Raman, industry analyst.

Currently, the Anthrop maintains its technical edge with the expansion of Claude’s symbol subscriptions to diversify beyond the dependency of API. The company’s ability to maintain its leadership will be determined by the next wave of competition from Openai and Google and others whether the rapid growth path continues or faces important opposite winds.

The risks cannot be higher: everyone who controls the tools of artificial intelligence that may ultimately control the development of energy programs at the pace of technological progress itself. In the last Silicon Valley battle, Anthropor built an empire on two clients-and now it must prove that it could keep it.