Our Large Quantitative Models

AI models grounded in the physical sciences and built for real-world impact.

Model Catalogue

Material Discovery

AQCat Adsorption Spin

AQCat Adsorption Spin enables researchers to lock in the first critical step of catalyst discovery and focus modeling and lab resources only on the most promising candidates.

It delivers near-DFT accuracy thousands of times faster, enabling cost-effective high throughput screening for the first time.

What is the binding energy of nitrogen (N2) on an iron (1,1,1) surface? Please find the most stable binding site and provide the relaxed structure.
Screen a set of alloy compositions and identify which candidates push binding energy toward the target threshold for this reaction.
Find surface compositions that improve selectivity for reactions such as carbon capture or nitrogen fixation.
Rapidly evaluate a large set of heterogeneous catalysts or organometallic complexes and rank the most promising candidates for follow-up.
Compare adsorption behavior across multiple metals, alloys, or surface facets and highlight the strongest candidates.
Identify which catalyst candidates we should prioritize first before committing DFT and lab resources.
Drug Discovery

AQPotency

AQPotency enables researchers to virtually screen large compound libraries, predict binding potency against specific targets, and run reverse screens across known off-targets to surface early safety risks.

It delivers potency predictions with uncertainty and applicability domain assessments, helping teams reduce costly wet-lab screening and focus resources on the most promising compounds.

We're spinning up a new EGFR mutant-selective program. Biology has the mutant assay running, but our wild-type EGFR counter screen isn't enabled yet. Can you predict wild-type EGFR potency for these compounds in the meantime and tell us which of those predictions we can actually trust? Compounds are in seriesA_compounds.csv and seriesB_compounds.csv
Predict pIC50 for this ALK (Q9UM73) analog and report uncertainty alongside the mean. Also, please run a kinome scan against this to assess off-targets. SMILES: Cc1ccc(c2csc(Nc3ccc(N4CCOCC4)cc3)n2)cc1C
Scan this JAK3 (P52333)-directed scaffold against the human kinome panel. I want the full off-target liability profile, not just a summary. SMILES: C[C@@H]1CCN(C[C@@H]1N(C)C2=NC=NC3=C2C=CN3)C(=O)CC#N
I want to test this new experimental feature in AQPotency. Please run the Bowes safety panel on my mTOR (P42345) lead and rank the panel hits by potency_mean descending so I can see the most likely worst offenders first. SMILES: Cc1c(-c2ccccc2)nc2nc(N)nc(-c3ccc4[nH]c(=O)oc4c3)n12
Material Discovery

AQCat Dispersion

AQCat Dispersion will expand SandboxAQ's catalyst modeling capabilities to large, commercially vital molecules, allowing researchers to computationally screen catalysts across a wider range of reactions, faster and at a fraction of the cost of lab experimentation.

Drug Discovery

AQCell

AQCell will enable researchers to computationally simulate how living cells respond to drug perturbations across thousands of candidates, predicting pathway-level efficacy and a "toxicity signature" of apoptosis pathway activation.

Material Discovery

AQCat Adsorption Spin

AQCat Adsorption Spin enables researchers to lock in the first critical step of catalyst discovery and focus modeling and lab resources only on the most promising candidates.

It delivers near-DFT accuracy thousands of times faster, enabling cost-effective high throughput screening for the first time.

What is the binding energy of nitrogen (N2) on an iron (1,1,1) surface? Please find the most stable binding site and provide the relaxed structure.
Screen a set of alloy compositions and identify which candidates push binding energy toward the target threshold for this reaction.
Find surface compositions that improve selectivity for reactions such as carbon capture or nitrogen fixation.
Rapidly evaluate a large set of heterogeneous catalysts or organometallic complexes and rank the most promising candidates for follow-up.
Compare adsorption behavior across multiple metals, alloys, or surface facets and highlight the strongest candidates.
Identify which catalyst candidates we should prioritize first before committing DFT and lab resources.
Material Discovery

AQCat Dispersion

AQCat Dispersion will expand SandboxAQ's catalyst modeling capabilities to large, commercially vital molecules, allowing researchers to computationally screen catalysts across a wider range of reactions, faster and at a fraction of the cost of lab experimentation.

Drug Discovery

AQPotency

AQPotency enables researchers to virtually screen large compound libraries, predict binding potency against specific targets, and run reverse screens across known off-targets to surface early safety risks.

It delivers potency predictions with uncertainty and applicability domain assessments, helping teams reduce costly wet-lab screening and focus resources on the most promising compounds.

We're spinning up a new EGFR mutant-selective program. Biology has the mutant assay running, but our wild-type EGFR counter screen isn't enabled yet. Can you predict wild-type EGFR potency for these compounds in the meantime and tell us which of those predictions we can actually trust? Compounds are in seriesA_compounds.csv and seriesB_compounds.csv
Predict pIC50 for this ALK (Q9UM73) analog and report uncertainty alongside the mean. Also, please run a kinome scan against this to assess off-targets. SMILES: Cc1ccc(c2csc(Nc3ccc(N4CCOCC4)cc3)n2)cc1C
Scan this JAK3 (P52333)-directed scaffold against the human kinome panel. I want the full off-target liability profile, not just a summary. SMILES: C[C@@H]1CCN(C[C@@H]1N(C)C2=NC=NC3=C2C=CN3)C(=O)CC#N
I want to test this new experimental feature in AQPotency. Please run the Bowes safety panel on my mTOR (P42345) lead and rank the panel hits by potency_mean descending so I can see the most likely worst offenders first. SMILES: Cc1c(-c2ccccc2)nc2nc(N)nc(-c3ccc4[nH]c(=O)oc4c3)n12
Drug Discovery

AQCell

AQCell will enable researchers to computationally simulate how living cells respond to drug perturbations across thousands of candidates, predicting pathway-level efficacy and a "toxicity signature" of apoptosis pathway activation.

Ready to try it out?

Start with $2,000 on Us  →