AI research lab Anthropic’s AI models are now the top choice for enterprises, surpassing OpenAI.
Anthropic now holds 32% of the enterprise large language model market share by usage, according to a report from Menlo Ventures released on Thursday. OpenAI holds the second-largest market share by usage among enterprises, with 25%.
The figure marks a strong reversal from even just a couple of years ago. Since 2023, OpenAI has seen its market share among enterprises decline sharply, according to the report, as Anthropic’s has steadily risen over the same timeframe. OpenAI held 50% of the enterprise market share by usage just two years ago while Anthropic had 12%.
Google has seen enterprise usage for its models increase over the last few years as well.
Anthropic has an even larger market share when it comes to coding, with 42% of the enterprise market share, the largest market share by a wide margin. Enterprise usage of Anthropic’s AI models are more than double OpenAI’s, when it comes to coding, which garnered 21% of overall market share.
Anthropic’s release of its Claude 3.5 Sonnet model in June 2024 is what laid the foundation for the company’s surge in usage, according to the report. The release of Claude 3.7 Sonnet in February 2025 only accelerated that momentum.
The findings from Menlo Ventures align with anecdotal chatter in the industry, which suggested that enterprise and startup developers preferred Claude over OpenAI’s ChatGPT. Meanwhile, OpenAI has a strong foothold on the consumer side of the house. The company reported last week that its users send more than 2.5 billion prompts to ChatGPT a day.
Techcrunch event
San Francisco
|
October 27-29, 2025
The Menlo Ventures report found enterprises prefer closed models, which Anthropic and OpenAI use. More than half of enterprises replied that they don’t use open source models at all. Only 13% of enterprise daily workloads use open source models as of mid-year 2025, down from 19% at the beginning of the year. Meta still maintains dominance in the open source market.