The 2025 Hype Cycle for GenAI Highlights Critical Innovations

Large language models (LLMs), which are pretrained on data, remain the cornerstone of GenAI and the most mature model technology on the Hype Cycle. These general purpose foundational models can be customized for a wide variety of use cases and offer significant capabilities, hence their prominence. However, other types of models, such as open-source LLMs, domain-specific GenAI models and large reasoning models, are quickly evolving into viable options for organizations. 

Example technology: Multimodal generative AI

For stronger, faster AI outcomes, use Gartner’s proprietary AI Use Case Insights tool to explore, evaluate and prioritize over 500 proven AI use cases tailored to your industry.

As organizations prepare to scale GenAI programs, the ability to build, govern and customize GenAI-powered applications becomes critical. AI engineering encompasses a growing ecosystem of GenAI tools and techniques that enables organizations to ensure Gen-AI-powered applications support their organization’s broader strategy. These tools offer effective application orchestration frameworks, reduce hallucinations, mitigate disinformation and ensure regulatory compliance. 

AI engineering also includes a group of tools and techniques focused on enforcing safe and efficient use of AI. 

Example technology: AI TRiSM

GenAI virtual assistants — e.g. ChatGPT — leverage LLMs to deliver functionality beyond that of traditional conversational AI technology and represent the most well known examples of GenAI in use today. 

Long-term, organizations want to use AI agents to automate complex, multistep processes at scale to boost productivity, lower operational costs and improve the customer experience.

Agentic AI, which is seeing a rapid surge of interest, autonomously or semi-autonomously uses AI techniques to perceive, make decisions, take actions and achieve goals in digital or physical environments. The shift from passive AI chatbots toward agentic AI marks a fundamental change in how organizations interact with AI systems and extract business value. 

Example technology: Embodied AI

GenAI’s evolution depends on a combination of novel techniques and established, foundational AI practices. Innovations like self-supervised learning, which reduces the need for large amounts of labeled training data, provide solutions to practical problems in the GenAI space. Currently, self-supervised learning is mainly involved in use cases like autonomous driving and medical diagnosis, but a growing number of industries have begun experimenting with the technology.

Specialized infrastructure is also gaining interest for its role in model training and the inference process. Specialized AI chips and tools can help increase efficiency and reduce associated costs.

Example technology: AI supercomputing