Turing says STEM AI agent now used at 415 U.S. universities

By AI, Created 5:20 PM UTC, May 18, 2026, /AGP/ – Turing says GPAI, its STEM-focused AI agent, is now in use at 415 U.S. universities, including MIT, Stanford and Harvard, as U.S. users and revenue have surged. The company is pushing deeper into research and enterprise workflows with new visualization tools, CBIT integration and Team and Enterprise plans.

Why it matters: - Turing is positioning GPAI as a purpose-built alternative to general AI tools in STEM fields where math, figures and technical accuracy matter. - The company says adoption across major U.S. universities and research users shows demand for tools that can handle research workflows end to end. - Rapid user and revenue growth suggests Turing is moving from an education product toward a broader research and enterprise business.

What happened: - Turing said GPAI is now used at 415 U.S. universities. - The list includes MIT, Stanford, Harvard, UC Berkeley, UCLA, the University of Michigan, UT Austin, Cornell, UIUC, Carnegie Mellon, the University of Pennsylvania and Columbia. - Turing said its cumulative U.S. user base grew 10x between September 2025 and April 2026. - The company said U.S. monthly active users grew at the same pace. - Turing said more than half of its global paying subscribers are in the U.S. and Europe. - The strongest concentration is among graduate students, postdoctoral researchers and faculty at research universities.

The details: - GPAI was designed for multi-step mathematics, physics derivations, structured data and publication-quality figures. - Its AI Solver works through problems in discrete reasoning steps, combining chain-of-thought with deterministic computation. - Its AI Visualizer creates diagrams and charts meant for journal submissions and slide decks. - AI Chat & Report Writer and AI Notebook support multi-model research, report generation from chat, and analysis of PDFs and video transcripts in one workflow. - The platform orchestrates multiple AI models so text, equations and visuals are processed together. - Turing says that orchestration helps GPAI produce quantitatively accurate visualizations instead of approximations. - A researcher at a U.S. AI lab said the AI Visualizer cut figure prep from about three hours to ten minutes. - A Ph.D. candidate in civil and industrial engineering at the University of Michigan said GPAI is now a daily tool for research, diagram generation and coding support. - An AP Physics high school teacher said step-by-step solutions and visualizations cut lesson-prep time and improved material quality. - The research behind GPAI was accepted to the industry track of EMNLP 2025. - The paper introduces CBIT, a framework for generating mathematics problems with large language models. - The company said CBIT showed lower error rates than human-written problems. - Turing said that is the first peer-reviewed evidence it has presented showing LLM-generated math content can meet production-grade accuracy in education. - Following paid plans, Turing said global revenue grew 830% over two months. - Turing offers Individual, Team and Enterprise plans. - The company has begun signing Team plan contracts with universities and corporate R&D groups. - Turing also supplies 300,000 math AI-ready training records to companies in Korea’s National Champion AI program.

Between the lines: - GPAI’s pitch is narrower than general-purpose chatbots and broader than a single-purpose tutor. - The product strategy centers on reducing friction in STEM work: solving, visualizing and writing in one place. - University adoption and paid-user growth suggest the most valuable use cases are moving from casual experimentation to recurring workflows. - The CBIT research gives Turing a technical proof point for AI-generated math content, which could help sales in education and research settings.

What’s next: - Turing plans to integrate CBIT into GPAI in 2026 as a built-in feature. - The company has already added GPT Image 2.0 to GPAI alongside Nano Banana Pro, Nano Banana 2 and Turing’s own STEM visualization engine. - Turing said it will keep integrating new image models as they emerge to improve equations, diagrams and data visualizations. - The company is expanding from B2C into B2B through graduate programs, research labs and enterprise R&D teams. - Turing’s longer-term goal is for GPAI to become a default AI agent for engineering, applied science, statistics and medicine.

The bottom line: - Turing is betting that STEM users will pay for AI that is built for technical work, not generic writing.