TPC leaders held discussions and an open plenary at TPC25 to identify potential TPC-wide initiatives, each to be fleshed out and evaluated with the intention of driving them forward through TPC quarterly hackathons beginning in January 2026 at SCA/HPCAsia26 in Osaka. TPC leaders seek input and involvement from the community regarding (a) individual and institutional interest in contributing to the refinement, planning, and pursuit of one or more of these initiatives, and (b) identification of alternative or additional initiatives.
Data & Shared Compute Infrastructure
Collectively build a scientific training data resource and shared infrastructure that can be used for model training and fine-tuning (general purpose FM, domain-specific FM, etc.)
Open Frontier Model
Build a frontier-scale, open AI model (using data from 1), harnessing the computational and data resources at many TPC partner institutions
Open Frontier AI system
Drive development and creation of Frontier AI system for science that incorporates reasoning models (start with SOA closed models, eventually include Open frontier model) and develops domain foundation models, knowledge graphs, orchestrations, simulators and experiments.
Software Infrastructure/Framework
Develop software infrastructure and middleware for building complex frontier-scale AI models and systems (This is in support of 2 & 3, also integrating experimentation/labs/instruments and other real-world connections)
Driving Challenge Applications
Identify challenge applications for driving and evaluating (1) - (3). Engage the community to volunteer and drive a diversity of scientific and industrial challenge applications that test and benefit from an Open Frontier AI System.
Open Suite for Evaluating Models for Science
Develop an open suite of tools, methods, and benchmarks for evaluating the scientific skills, knowledge, reasoning, agentic capabilities and safety/security of Frontier Models and AI systems.
