VIDraft

Foundation models, on-device AI, and computational science. We publish what we build.

What is here

Language and decision models, quantization and compression work, and applications that run from servers down to phones.

Open science, free to enter

We design and operate free public benchmark competitions so that research results can be contested and verified by anyone against the same standard. Entry is free, prize money is self funded, and the data and scoring rules are published.

See the FINAL-Bench organization for the open challenges in drug discovery, materials, superconductors and quantum computing.

Follow the work

• LinkedIn: VIDraft
• GitHub: final-bench