From drug discovery to drug design.

Ten63 brings Superlearning to drug discovery: pairing generative chemistry with quantum-accurate simulation to design therapeutics for currently intractable diseases.

And new data is brutally slow.

Every hypothesis has to be physically made and tested. Synthesis and assays are slow and costly, so the data that would unlock the rest barely trickles in.

So we made the simulation indistinguishable from reality.

Generate the data in silico, learn from it instantly, and the frontier moves. Superlearning turns the undruggable into the druggable.

Learninterpret resultsinstant feedbackTestassay in the labMakephysically synthesizeSimulateno synthesis, no assayDesignpropose a moleculeTHE BOTTLENECKmaking + testing isslow and expensiveIN SILICOtest any hypothesis,explore 100T molecules

Most of biology is still dark.

Models are only as good as their data, and the best of it is a vanishing slice of chemistry. Roughly 80% of the human proteome stays undruggable.

Step inside the model where chemistry becomes computable.

BEYOND, the world's first Large Quantum Chemistry Model.

Quantum-physics levels of accuracy at the speed of classical molecular mechanics. BEYOND virtually explores more than 100 trillion molecules, guided by a simulation environment that closely tracks experimental reality, to design optimal solutions.

>0%*
more accurate
>0×*
faster to compute
>0×*
cheaper to train

* Internal benchmarks vs Meta UMA-M.

BEYOND is the most accurate simulation.

External benchmark · GMTKN55 leaderboard ↗
0.1110100Mean absolute error vs CCSD(T), kcal/mol (lower is better)BEYONDMeta's UMA-MOther competitorsS22 (44)S66 (44)ADIM6 (44)WATER27 (15)
Mean absolute error vs CCSD(T) on public GMTKN55 subsets (S22, S66, ADIM6, WATER27). BEYOND matches or beats Meta's UMA-M and sits well below the distribution of all other competitor MLFFs (number of MLFF tested).
Internal benchmark
0246810Error vs gold-standard QM (lower is better)experimental accuracyGAFF2SQM2.20Meta's UMA-MBEYOND
Error distribution across ~5,600 drug-like and protein-like systems calculated at CCSD(T)/CBS. BEYOND is significantly more accurate than Meta's UMA-M (P < 0.0001, KS test.)

BEYOND achieves state-of-the-art accuracy at unprecedented speed.

0.010.11101001000MD steps / second · higher is betterBEYONDTen63260.3UMA-SMeta0.144≈ 1,808× slowerUMA-MMeta0.0188≈ 13,876× slower
Molecular-dynamics throughput on identical hardware (Google Cloud a2-ultragpu-2g, 24 vCPUs, 340 GB, 1×A100 GPU). Higher is better; log scale.

Only BEYOND breaks the trade-off frontier.

Until now, simulation forced a choice: accurate or fast. Highly accurate methods would take millions of years; fast methods have little predictive power. BEYOND clears both the quantum-accuracy and simulation-efficiency barriers at once.

Accuracy (1 / error vs gold standard) →Speed (1 / run time) →former trade-off frontierQuantum accuracy barriererror < 0.2 kcal/molSimulation efficiency barrier (< 0.05 s/calc)GAFF2SQM2.20Meta UMABEYONDLQCM

A clear path to drug-discovery superintelligence.

BEYOND creates the perfect environment for Superlearning. Industry-leading generative chemistry explores over 100 trillion molecules per pocket, while quantum-accurate simulations enable in silico experiments that are nearly indistinguishable from reality, at unprecedented scale. Each makes the other smarter, compounding toward superintelligence.

Generative chemistry
novel molecules designed across a vast space, at scale
×
Simulation
evaluated at quantum accuracy, indistinguishable from reality
=
Superintelligence
a clear, compounding path no one else can take

Breakthroughs against previously undruggable and under-drugged targets.

Every program starts with intelligence at scale. Sentinel, our virtual CSO, continuously scans all publicly available information and proposes targets ranked by confidence and weighted by patient need. Sentinel establishes a translational plan and identifies killer experiments before BEYOND ever designs a molecule.

Sentinel
Scan the field
literature, clinical & competitive signal, continuously
Sentinel
Target theses
ranked by confidence, each with an experiment that could disprove it
BEYOND
Design + simulate
generative chemistry, quantum-accurate simulation
Lab
Validate
synthesis & assays, GO / NO-GO gates
Outcome
Pipeline
first-in-class & best-in-class programs

The result: an internal pipeline of first-in-class and best-in-class programs across oncology and other serious diseases, advancing toward a future where no loved one is lost to disease.

An interdisciplinary team driving our mission.

Founders and operators from Duke, Pfizer, Genentech, and Mirati, uniting chemistry, physics, biology, and AI.

Marcel Frenkel, PhD
Marcel Frenkel, PhD
Founder & CEO
Mark Hallen, PhD
Mark Hallen, PhD
Founder & CTO
John Ketcham, PhD
John Ketcham, PhD
VP & Head of Chemistry
Charlie Holst, PhD
Charlie Holst, PhD
VP & Head of Biology
Nate Guerin, PhD
Nate Guerin, PhD
VP & Head of Engineering

Our investors.

Strategic and venture investors committed to building a better future for patients.

Ten63 Therapeutics