The Pulse
Iambic Unveils 41B-Parameter Enchant v3 for Drug Discovery
Iambic announced Enchant v3, a 41-billion-parameter multimodal transformer designed for end-to-end drug research and development. The company says the model adds new prediction capabilities and supports its internal and partnered discovery
AI.info Team ·
Iambic unveiled Enchant v3 on September 21, 2026, describing it as a 41-billion-parameter multimodal transformer designed to support drug discovery from early research through clinical development. The San Diego company says the model is the core artificial-intelligence technology powering its molecular superintelligence platform.
The launch marks a substantial increase in model size for Iambic. Enchant v1, trained in 2024, had 1 billion parameters, while Enchant v2, trained in 2025, reached 7 billion parameters. Enchant v3 adds a mixture-of-experts architecture, increased data scale, a broader range of data modalities and new prediction capabilities intended to test multiple drug-discovery hypotheses in parallel.
From 1 Billion to 41 Billion Parameters
Iambic describes Enchant as the central AI system in its molecular research platform. The model combines information from different types of biomedical data to estimate how a candidate compound may perform across a target profile. The system is designed to ingest molecular and biomolecular structures, sequence and omics data, tabular assay data, physics data, biomedical text and images, then produce predictions across preclinical and clinical endpoints.
Rather than treating each stage of drug discovery as a separate technical problem, Iambic says its platform is designed to address the process from hit identification through multiparameter lead optimization and clinical developability. Enchant works alongside automated laboratory data generation, with predictions helping guide design choices and prioritize experiments.
Those predictions include uncertainty estimates, which the company says help determine which laboratory experiments should be run next. Iambic reports that its teams fine-tune the model each week on hundreds of distinct molecular properties and generate approximately 12 million inferences per month across wholly owned and partnered drug-discovery programs.
The company also says Enchant can transfer information between related endpoints, such as using less expensive in-vitro measurements to help predict more expensive in-vivo properties, or using preclinical information to estimate clinical characteristics.
New Predictions for Molecules, Omics and Peptides
Iambic says Enchant v3 adds inverse design, omics-based prediction, image-based prediction and support for peptides. Those capabilities extend the model beyond conventional small-molecule property prediction and support the company’s stated aim of applying one system across hit identification, lead optimization and clinical development.
Iambic reports that prediction accuracy has improved as the Enchant models have grown. The company’s comparison of the three generations shows performance improving with increased scale across a range of preclinical and clinical endpoints.
IAM-C1 and IAM217 Supply the Drug-Discovery Test Cases
Iambic points to two internal programs as examples of how Enchant is used alongside automated chemistry and biology experiments. IAM-C1 is a selective dual inhibitor of CDK2 and CDK4, designed to inhibit two cancer-related pathways while avoiding closely related kinases that could narrow its therapeutic window. The company says the candidate has four off-targets in a KINOMEScan TREEspot analysis.
The second example, IAM217, is an allosteric KIF18A inhibitor being studied for triple-negative breast cancer, ovarian cancer and other solid-tumor indications. Iambic reports that IAM217 produced approximately 90 percent regression of established intracranial tumors in preclinical studies, while a presumed clinical-stage comparator produced no regression. The company also says IAM217 achieved comparable tumor-growth inhibition at lower total and unbound plasma exposures than the standard of care and demonstrated a clean drug-to-drug interaction profile supporting once-daily oral dosing.
Those findings are company-reported preclinical results, not evidence of clinical benefit. Iambic says both IAM-C1 and IAM217 are advancing toward submissions for investigational new drug applications, with human clinical data still required to establish whether the programs translate.
Deploying Enchant Across Drug Research
Iambic says it plans to deploy Enchant v3 across its internal and partner drug-discovery work. The company expects the model to help test multiple hypotheses, predict preclinical and clinical endpoints and prioritize potential compounds for laboratory experiments.
For now, the public evidence consists of Iambic’s reported model comparisons, platform description and preclinical examples. The company says final proof will come from human clinical data as its programs advance toward the clinic.