Pilot for Drug Discovery Labs

High-Accuracy Quantum Chemistry Labels

Your molecules, labelled at CCSD(T) grade and tokenised as quantum inputs for your model. Faster to make, and your model trains on far less data.

Accurate< 1 kcal/mol

Within chemical accuracy of CCSD(T)/CBS-level reference labels.

Fastup to 800×

Faster than the standard CCSD(T)-level label recipe.

Label-efficient100-340×

Fewer labels than a standard descriptor model for the same error, on the same CCSD(T)/CBS-level label and held-out molecules. Chemical accuracy where the descriptor model stops at 1.8 kcal/mol.

Request a pilot

Built for drug discovery teams

AI drug discovery teams

Train your models on gold-standard data and rank compounds you can trust, with far less data.

Computational chemistry

Gold-standard numbers for the molecules where DFT and force fields go wrong.

Discovery programmes

Answers in days, not months, so your design cycle never waits on quantum chemistry.

Already in researchers' hands

3,800+

Downloads of our open quantum datasets on Hugging Face.

See the datasets →
99.6%

Of held-out molecules within chemical accuracy with Canis M, our open model.

Try Canis M →
320×

Fewer labelled molecules than a standard model, on our FCI-label benchmark. Published on ChemRxiv.

Read the paper →

Methods we support

CCSD(T)

Gold-standard accuracy for small and drug-fragment-sized organic molecules.

FCI

Exact reference data for small systems.

DMRG

Strongly correlated systems, such as metal centres. On request.

ANI-1ccx / ANI-2x

Drop-in format for your existing training pipeline.

Your molecules. Gold-standard labels. In days.

Send us a molecule set. We return it labelled and tokenised, with a report against the labels you use today.

Request a pilot