Chai Discovery vs EvolutionaryScale (2026)
Both build foundation models for protein work and make them available to other organizations. Chai Discovery licenses design models to pharmaceutical and life sciences teams: Chai-2 and Chai-3 design antibodies and miniproteins from a named target and epitope, and Chai-1 was released for structure prediction with open weights. EvolutionaryScale, now part of Chan Zuckerberg Biohub, publishes the ESM family: ESM3 models sequence, structure and function together and was published in Science in 2025, and the models are served on the Biohub Platform with a free credit tier. Chai is aimed at binder design for drug programs; ESM is a general protein model family that researchers can prompt for many tasks.
- De novo design of antibodies and miniproteins from a target and epitope, including monoclonal and VHH formats.
- Published results per target class with counts, and an open weights structure model, Chai-1.
- Models licensed to research organizations rather than used for an internal pipeline.
- ESM3, which takes any combination of sequence, structure and function as a prompt, published in Science in 2025.
- Models served through the Biohub Platform with a free credit tier, and model cards that state intended use and clinical limits.
- A model family that extends to ESM Cambrian for representations and ESMFold2 for structure prediction.
This comparison is published by AI Health Index, an independent research platform that compares healthcare AI vendors objectively. Chai Discovery and EvolutionaryScale are each graded against the same capability taxonomy, from each vendor's own public materials and the regulatory record, under the AI Health Index verification standard. No vendor pays for placement, and no vendor has reviewed this page. How this evidence is graded
Plain facts
| Fact | Chai Discovery | EvolutionaryScale |
|---|---|---|
| Primary category | Drug Discovery AI | Drug Discovery AI |
| Founded | 2024 | Not recorded |
| Headquarters | San Francisco, California | New York, NY |
| Website | chaidiscovery.com | evolutionaryscale.ai |
Side by Side
Select any grade to read the note behind it.
Product updates
The latest logged changes from each vendor, newest first, each linked to its source.
Chai Discovery
-
Model / architecture
Chai Discovery deployed Chai-3, a next generation antibody design model that the company describes as roughly doubling the success rate of Chai-2 and binding more tightly to harder targets.
Source
Each vendor in one paragraph
Written to be quoted whole. Each paragraph opens the AI Health Index profile of the vendor: what the product is, who buys it and what it does.
San Francisco company that licenses foundation models for molecular design to pharmaceutical and life sciences research organizations, rather than using them to build its own pipeline. Three generations are published: Chai-1 (2024) for biomolecular structure prediction, released with open weights under a community license; Chai-2 (2025) for de novo design of antibodies and miniproteins from a target and epitope prompt; and Chai-3 (2026), which the company describes as roughly doubling its predecessor's success rate.
Source: AI Health Index, September 2026
AI research lab building protein language models, founded by researchers who developed the ESM family at Meta and now part of Chan Zuckerberg Biohub, which serves the models and their documentation on the Biohub Platform under Biohub's terms. ESM3 models a protein's sequence, structure and function together, so a scientist can supply any combination of the three as a prompt and generate a protein with specified properties; it was published in Science in January 2025 and trained on 2.78 billion protein sequences at 98 billion parameters.
Source: AI Health Index, September 2026
Questions buyers ask
Which settings and specialties do Chai Discovery and EvolutionaryScale cover?
For Chai Discovery, the company does not claim clinical or downstream development capability. Testing runs across the classes it names: 28 target antigens in the developability work, six G protein coupled receptors, a class the company notes is hard for screening methods, and three peptide MHC complexes, each with the number of designs tested reported.
With EvolutionaryScale, the models work on proteins as a class rather than on a disease area, and the company names the uses it is built for: designing enzymes that break down plastic, proteins that capture carbon, antibodies, and new medicines, with the ESM Atlas covering proteins across all of life. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified September 28, 2026. No vendor pays for placement.
What evidence is published for Chai Discovery and EvolutionaryScale?
In Chai Discovery's case, the product page states double digit success rates for antibodies and above 50 percent for miniproteins, and characterization within two weeks without high throughput screening.
For EvolutionaryScale, the platform's scale is visible in the ESM Atlas, an open database of annotations for more than 6.8 billion proteins with 1.1 billion predicted structures. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified September 28, 2026. No vendor pays for placement.
How much do Chai Discovery and EvolutionaryScale leave to the clinician?
With Chai Discovery, access is controlled rather than open, granted on application under a stated deployment policy, and an acceptable use policy binds what may be done with the models, their outputs and derivatives, with suspension for violation and a prohibition on attempts to circumvent safety filters.
In EvolutionaryScale's case, the model cards and the Limitations page state that outputs are computational predictions, not experimentally validated, and not intended for clinical or therapeutic use without further validation, and the ESM Atlas search is flagged as using an agent that may answer inaccurately. Nothing acts on its own: a design is a proposal until the user expresses and characterizes it. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified September 28, 2026. No vendor pays for placement.
Related comparisons
Other published head to head assessments involving these vendors or their closest peers.
Both records were reviewed on this site in September 2026. EvolutionaryScale's models now run under Biohub's terms.
Ask both how they report designs that failed in the lab alongside those that worked.