- The developers go from loyalty to flexibility at the idea of Openai, but Deepseek quickly wins ground
- Microsoft has trouble for the Mindshare developer with only an adoption rate of 10%
- The integration of AI becomes leaner as the teams deposit formal structures and focus on usable tools
Openai has dominated the LLM space for some time and remains the main choice for many developers, but market dynamics and loyalty seem to change, said new research.
A Vercel survey of 656 application manufacturers revealed that 87% use OPENAI models and 83% are based on its inference APIs. However, developers now use two suppliers on average and 60% have changed sellers in the past six months – raising questions about their loyalty.
Deepseek is now used by almost a third of companies (29%), while Microsoft’s LLMs, on the other hand, seem to have trouble, with only 10% of respondents reporting by using them, and 9 out of 10 developers do not consider it as a viable option.
“AI dissolves the boundaries between roles. We see new products of products mixing UX, UI and code in a creative flow – thanks to tools like Vercel, V0, UIZARD and Cursor. Whether it is the junior or senior leader, anyone can now build, test and send ideas independently – and it is not only effective, it is liberating, “said Nicolas Le Plec, CTO, Emea – Akssa.
AI is now built around clear use cases, hierarchical developers or tools – not teams. Forty-five percent of respondents said they did not have a dedicated AI team, while 57% said any specific AI leadership structure. Instead, success with AI depends on clear priorities and good tools.
“By kissing advanced AI technologies, we allow our teams to work more intelligently and faster,” said Dr. Jan Ittner of BCG X, echoing the feeling that a writer or AI developer tool can be more precious than another hire.
The features of the AI of products are now a priority for 75% of customer -oriented applications, while only 39% still include traditional support chatbots. The personalization of the website remains underused at 27%, which indicates a place for future innovation.
Cost control is also a major priority. More than 70% of developers manually test their models, spending less than $ 1,000 per month – showing how much can be made on limited budgets. To reduce costs, only 14% form their models, while 60% use the generation of recovery (RAG) and vector databases.
The emphasis on the tools stems from an “superhype” perceived around the AI, with an average note of 6.4 out of 10. However, the developers gave a 7.7 out of 10 when asked if the AI would transform their industries over the next 12 months.
In a changing field where speed, precision and adaptability count more than the scale, the best LLM for coding is not corrected – it is the one that solves the problems of today while keeping the door of tomorrow open.
