Frontier AI Benchmarking Datasets

The Frontier AI Benchmarking Datasets competition is designed to strengthen the UK’s position in the global AI landscape by funding projects that build the data infrastructure needed for next-generation AI systems. The call emphasizes on creating high-quality datasets and benchmarking frameworks that enable consistent evaluation, validation and training of frontier AI models.

Projects are expected to deliver both benchmarking tools and full datasets, ensuring that outputs are not only technically robust but also usable by industry and researchers. A strong focus is placed on FAIR data principles and ensuring that datasets are structured, annotated and documented in a way that maximises reuse and long-term value.

Funding amount

Between £500,000 and £750,000

What your project must deliver

• To be competitive, proposals need to go beyond simple dataset creation and demonstrate a complete ecosystem for AI benchmarking and utilisation. This includes:
• Your project must produce an open benchmark package, including clearly defined benchmark tasks, evaluation protocols, and an evaluation harness that can be used by third parties. The benchmark dataset itself must be accessible and accompanied by detailed documentation and metadata.
• Projects must clearly outline how the dataset will be used beyond the project, including IP ownership, licensing models and routes to industry access. Innovate UK is looking for proposals that demonstrate how these assets will create long-term commercial and strategic value.

Thematic focus areas

Your proposal must align with one of two priority themes; each linked to national strategic goals:

AI-Enabled Health and Life Sciences

Projects in this theme should support AI applications across areas such as drug discovery, clinical trials, predictive healthcare and manufacturing. Where relevant, enabling technologies such as genomics and multi-omics can be incorporated.

Advanced Materials with AI

This theme focuses on using AI to accelerate materials discovery and optimisation across sectors such as aerospace, semiconductors and net zero technologies. Projects may explore generative models, physics-based machine learning or multimodal data platforms.