ChatGPT's market share has dropped below 50% for the first time since its launch, marking a significant shift in the AI assistant landscape.

Users are migrating to competing platforms including Google's Gemini, Anthropic's Claude, and xAI's Grok. The shift signals growing willingness among users to experiment with different AI assistants rather than sticking with a single platform.

Despite the decline, ChatGPT remains the world's most popular AI assistant. OpenAI's flagship product was the fastest app ever to reach one billion monthly users.

Competition intensifies across the board

The market fragmentation reflects the maturation of the AI assistant space. Multiple providers now offer competitive capabilities, reducing the switching costs for users seeking different features or interaction styles.

Meanwhile, Cursor announced plans to release a massive new model trained from scratch on over 100,000 GPUs. The model contains more than 1.5 trillion parameters and aims to push "agentic software development far beyond autocomplete and pair programming," according to the company.

The model will launch in the coming weeks and represents Cursor's bid to expand beyond its current code editor focus into broader AI assistance.

Vercel also entered the agent infrastructure space with the launch of eve, an open-source framework designed for production AI agents. The platform includes durable execution, sandboxed compute, approvals, and evaluation capabilities.

Separately, OpenAI secured a notable hire with Noam Shazeer joining the company from Google. Shazeer co-authored the seminal 2017 "Attention Is All You Need" paper that introduced the transformer architecture underlying modern large language models.

The hiring represents a significant coup for OpenAI as it competes for top AI talent. Shazeer previously worked on Google's search spelling correction and created core algorithms for AdSense.

OpenAI also introduced LifeSciBench, a new benchmark evaluating AI systems on complete life sciences workflows rather than isolated biology tasks. The benchmark covers evidence analysis, experimental design, and research communication.