Key takeaways of zdnet
- Programming is AI’s Killer App.
- Top business AI, especially for programming, is anthropic.
- Open-source AI is behind its ownership contestants.
If you are on the road. Random users used to ask what is the most popular business AI Large Language Model (LLM), so I am sure they will say they will say OpeniChatgate. By mid -2012, however, Anthropic 32% is the leading enterprise LLM provider According to the use of the enterprise, according to Meno venturesAn early stage enterprise capital firm.
Before you are very excited, however, keep in mind that Menlo Ventures is a major anthropic investor. The firm has supported the company through several important funding rounds, including their series D Round and participating in their $ 3.5 billion series E, with anthropic priced $ 61.5 billion.
Also: What happened when the cloud AI of Anthropic AI launched a small shop for a month (Spiiler: It turned strange)
In other words, Menlo Ventures have billions of reasons for praising anthropic. He said, other people also see anthropic as the top venture AI company. As AI magazine said, “Anthropic has established itself as a premiere enterprise AI company Through its cloud family of LLM, achieving a 1,000% year-year increase in annual recurring revenue to reach $ 3 billion. “Even by hyper-invasive AI standards, this is real development.
Back anthropicYou will get Openai, which now has 25%; Google with 20%; And Meta Lama With 9%. All the way back, you will get with only 1% DeepsekAfter the rest of the pack. Menlo Ventures credited the rapid climb of anthropic for its strong performance Cloud sonnet And Cloud opas Model.
(Disclosure: ZDNET’s original company Ziff Davis filed a case of April 2025 against Openai, alleging that it violates Ziff Davis copyright training and operating its AI system.)
This number shows the ratio of the use of production AI, not spent. They were obtained from the survey of 150 technical decision making in enterprises and startups for the manufacture of AI applications in summer of 2025.
What is behind the success of anthropic?
Three separate factors are increasing the rise of anthropic. The first is that Menlo Ventures called “AI’s first killer app”: Code Generation. While the AI-made code quality remains suspicious, more, more developers are using AI programming tools more than ever, and the cloud has become the top choice of the programmer with 42% of the market share. It is 21% of double openi.
There are solid examples of anthropic development programs that gain popularity. For example, in only one year, Claude helped the conversion Gitab Copilot In a $ 1.9 billion ecosystem. The 2024 release of Cloud Sonnet 3.5 showed how LLM successes can make it possible to make it possible to fully make new categories such as AI IDES, Cursor and Windsurf; Vibe app builders, cute, bolt, and answer; And enterprise coding agents, cloud code and all hands.
Another reason is its use of anthropic winning Learning reinforcement with verification awards (RLVR) To train its LLM. There is a simple concept behind that complex name: you provide clear, binary feedback (1 for the right) on the output of the model. It works well for programming AI tools, where the code either works or not.
Anthropic has also led the route to LLM that takes a step-by-step approach to solve problems and use external devices to draw into data to give better answers. In short, anthropic AI has been leading to agents. In addition to helping people and programmers, this approach can help LLMS improve their reactions and integrate devices such as search, calculator, coding environment and other resources through model reference protocol (MCP). This new open-source protocol enables LLMS and AI agents to be originally able to connect with the huge, sometimes changing landscapes of real-world data, equipment and services.
Also: 7 strategic insights trade and IT leaders are required for AI change in 2025
This is important because Meno Ventures also found that it is not a price that motivates companies to replace LLM, it is a performance. “It makes an unexpected market dynamic: even 10x falls in the price of individual models, builders do not occupy savings using old models; they move N Mess N Mess for just the best performing one.”
Once LLM matures it can turn dynamic and models start reaching the same performance level. For now, however, as LLMs improve massively from one release to another, companies are ready to pay for the latest and fastest.
In the enterprise
The study also found that companies are constantly changing from manufacturing and training models to estimates, ie with the model actually running in production. Startups are proceeding in this way, with 74% of the builders now saying that most of their workloads are in production. Large enterprises are not far behind with 49% reporting that most or almost all of their computer are in production. In short, enterprises are now using AI, not only the construction of AI.
In the end, researchers said that open-source LLM today declined by 13% of AI workload by 19% six months ago. Market leaders remain Lama, although Lama is not a really open source.
Also: How the agent AI is changing the very foundation of business strategy
Nevertheless, more open-sources LLMs are visible. These include the new models of Deepsek (V3, R1), Bidens Seed (Dobao), Minimax (Text 1), Alibaba (Quven 3), Moonshot AI (Km 2), and Z. AI (GLM 4.5) in the last six months.
They are not just using much. This is because despite their advantages, “more adaptation, potential cost savings, and the ability to deploy within private clouds or on-rich environment,” their performance has continued “Trail Frontier, Close-SOS model”. Add it that many of the best performing open-sources are from LLM Chinese companies, which are careful of Western business, and open-sources LLM appear to stop outside.
Also: Anthropic made its revenue three times in 5 months – and this is the reason
So what does it all do?
Menlo Ventures said, “It can be a foolish to predict the future of AI. With the weeks change, the exciting new model launch, with the progress in foundation model capabilities and reducing the cost,” said Menlo Ventures. Nevertheless, “Today’s founder is mature for a new generation of AI businesses to be built on top of building blocks.”
However, the question is “What will be those basic building blocks?” Openai? Google, Meta? Anthropic? Stay We are not yet able to say that the AI model will eventually end at the top.
Get top stories of morning with us in your inbox every day Tech Today Newsletter.
