Anthropic Beats Openai As The Top Llm Provider For Business - And It's Not Even Close

oxygen/Getty

ZDNET's cardinal takeaways

• Programming is AI's slayer app.
• The apical business AI, particularly for programming, is Anthropic. 
• Open-source AI is lagging down its proprietary competitors.

If you were to inquire J. Random User connected nan thoroughfare what nan astir celebrated business AI Large Language Model (LLM) is, I stake you they'd opportunity OpenAI's ChatGPT. As of mid-2025, however, Anthropic is nan starring endeavor LLM provider, pinch 32% of endeavor usage, according to Menlo Ventures, an early-stage task superior firm.

Before you get excessively excited, though, support successful mind that Menlo Ventures is simply a awesome Anthropic investor. The patient has backed nan institution done respective important backing rounds, including starring their Series D information and participating successful their $3.5 cardinal Series E, which weighted Anthropic astatine $61.5 billion. 

Also: What happened erstwhile Anthropic's Claude AI ran a mini shop for a period (spoiler: it sewage weird)

In different words, Menlo Ventures has billions of reasons to praise Anthropic. That said, others besides position Anthropic arsenic nan apical endeavor AI company. As AI Magazine put it, "Anthropic has established itself arsenic nan premier endeavor AI company done its Claude family of LLMs, achieving singular 1,000% year-over-year maturation to scope $3 cardinal successful yearly recurring revenue." Even by hyper-aggressive AI standards, that's existent growth.  

Behind Anthropic, you'll find OpenAI, which now has 25%; Google pinch 20%; and Meta Llama with 9%. All nan measurement successful nan back, pinch a specified 1% you'll find DeepSeek, followed by nan remainder of nan pack. Menlo Ventures credits Anthropic's accelerated ascent to nan beardown capacity of its Claude Sonnet and Claude Opus models.

Menlo Ventures

(Disclosure: Ziff Davis, ZDNET's genitor company, revenge an April 2025 suit against OpenAI, alleging it infringed Ziff Davis copyrights successful training and operating its AI systems.)

These numbers bespeak nan proportionality of accumulation AI use, not spending. They were derived from a study of 150 method decision-makers astatine enterprises and startups building AI applications successful nan summertime of 2025. 

What's down Anthropic's success?

Three different factors are driving Anthropic's rise. The first is what Menlo Ventures calls "AI's first slayer app": Code generation. While AI-created codification value remains questionable, nevertheless, much developers are utilizing AI programming devices than ever, and Claude has go programmers' apical prime pinch 42% of nan marketplace share. That's double OpenAI's 21% share. 

There are actual examples of Anthropic improvement programs gaining popularity. For instance, successful conscionable 1 year, Claude helped toggle shape GitHub Copilot into a $1.9-billion ecosystem. Claude Sonnet 3.5's 2024 merchandise showed really LLM breakthroughs tin make imaginable wholly caller categories specified arsenic AI IDEs, Cursor and Windsurf; vibe app builders, Lovable, Bolt, and Replit; and endeavor coding agents, Claude Code and All Hands. 

Another logic Anthropic is winning is its usage of reinforcement learning pinch verifiable rewards (RLVR) to train its LLMs. Behind that analyzable sanction lies a elemental concept: You supply clear, binary feedback (1 for correct, 0 for incorrect) connected nan model's output. This useful good for programming AI tools, wherever nan codification either useful aliases doesn't.

Anthropic has besides led nan measurement to LLMs that return step-by-step approaches to solving problems and usage outer devices to propulsion successful information to present amended answers. In short, Anthropic has been a leader successful creating AI agents. Besides helping group and programmers, this attack tin thief LLMs iteratively amended their responses and merge devices for illustration search, calculators, coding environments, and different resources via nan Model Context Protocol (MCP). This caller open-source protocol enables LLMs and AI agents to seamlessly link pinch nan vast, ever-changing scenery of real-world data, tools, and services.

Also: 7 strategical insights business and IT leaders request for AI translator successful 2025

That's important because Menlo Ventures besides recovered that it's not value that drives companies to alteration LLMs, it's performance. "This creates an unexpected marketplace dynamic: Even arsenic individual models driblet 10x successful price, builders don't seizure savings by utilizing older models; they conscionable move en masse to nan best-performing one."

This move whitethorn alteration erstwhile LLMs commencement to mature and models statesman to scope akin capacity levels. For now, though, arsenic LLMs amended massively from 1 merchandise to another, companies are consenting to salary for nan newest and fastest. 

AI successful nan enterprise

The study besides recovered that companies are steadily shifting from building and training models to inference, that is, pinch models really moving successful production. Startups are starring nan way, pinch 74% of builders now stating that astir of their workloads are successful production. Large enterprises aren't acold behind, pinch 49% reporting that astir aliases astir each of their computers are successful production. In short, enterprises are now utilizing AI, not simply building AI.

Finally, nan researchers said that open-source LLMs person declined to 13% of AI workloads coming from 19% six months ago. The marketplace leader remains Llama, albeit that Llama isn't really unfastened source.

Also: How agentic AI is transforming nan very foundations of business strategy

Nevertheless, much open-source LLMs person been appearing. These see caller models from DeepSeek (V3, R1), Bytedance Seed (Doubao), Minimax (Text 1), Alibaba (Qwen 3), Moonshot AI (Kimi K2), and Z AI (GLM 4.5) successful nan past six months. 

They're conscionable not utilized much. That's because, contempt their advantages, "greater customization, imaginable costs savings, and nan expertise to deploy wrong backstage unreality aliases on-premises environments," their capacity has continued to "trail frontier, closed-source models." Add successful that galore of nan best-performing open-source LLMs to day are from Chinese companies that Western businesses are wary of, and open-source LLMs look to beryllium stalling out.

Also: Anthropic tripled its gross successful 5 months - and this is why

So what does each this lead to?

"Predicting nan early of AI tin beryllium a fool's errand. The marketplace changes by nan week, pinch breathtaking caller exemplary launches, advancements successful instauration exemplary capabilities, and plunging costs," Menlo Ventures said. Still, "conditions are ripe for a caller procreation of enduring AI businesses to beryllium built connected apical of today's foundational building blocks."

The mobility remains, however, "What will those foundational building blocks be?" OpenAI? Google, Meta? Anthropic? Stay tuned. We're not yet adjacent to being capable to opportunity which AI models will yet extremity up connected top.   

Get nan morning's apical stories successful your inbox each time pinch our Tech Today newsletter.