Meta AI ESM
Meta AI ESM (Evolutionary Scale Modeling) is a research project and open-source software library developed by Meta's Fundamental AI Research (FAIR) team, focused on applying large language model techniques to protein sequences.
Profile
Meta AI ESM provides open-source deep learning models that predict protein structure, function, and evolutionary relationships from amino acid sequences.
Meta AI ESM (Evolutionary Scale Modeling) is a research project and open-source software library developed by Meta's Fundamental AI Research (FAIR) team, focused on applying large language model techniques to protein sequences. The project, led by researchers including Alexander Rives, was first released in 2020 and gained significant traction in the computational biology community for its ability to predict protein structure and function from sequence data alone. The core repository, hosted on GitHub under facebookresearch/esm, was archived and set to read-only on August 1, 2024, indicating a shift in Meta's research priorities or project lifecycle.
Despite the archive status, the ESM models remain widely used, with the repository accumulating over 4,100 stars and 794 forks. Meta itself reported $200.97 billion in total revenue for 2025, with a net income of $60.46 billion, and as of March 31, 2026, employed 77,986 people. The company's capital expenditures for 2026 are forecast between $125 billion and $145 billion, heavily driven by AI infrastructure investments.
In April 2026, Meta released its first model from the newly formed Meta Superintelligence Labs, signaling a broader strategic pivot toward advanced AI systems. The ESM project, while not a standalone commercial product, represents Meta's contribution to open-source AI for the life sciences, competing with other protein language models like Google DeepMind's AlphaFold and ProtGPT2. The project's impact is measured through academic citations and downstream applications in drug discovery, enzyme engineering, and synthetic biology, rather than direct revenue generation.
Meta's broader AI strategy includes a reported multi-billion dollar investment in Scale AI in June 2025, potentially exceeding $10 billion, and the acquisition of Assured Robot Intelligence in May 2026 to build humanoid technology. The company faces ongoing legal and regulatory challenges, including youth-related lawsuits and EU regulatory headwinds, which could materially impact its financial results. ESM itself has no dedicated headcount or revenue, operating as a research output within Meta's broader AI research organization.
Who buys this
- Academic researchers in computational biology and bioinformatics
- Pharmaceutical and biotechnology companies conducting drug discovery
- Synthetic biology firms engineering novel enzymes and proteins
- AI researchers studying protein language models and their applications
- Open-source developers integrating protein prediction into scientific workflows
Strengths and what to watch
Strengths
- Open-source availability with permissive license (MIT) enables broad adoption and community contributions across academia and industry
- Proven accuracy in protein structure prediction, demonstrated through benchmarks like CASP and CAMEO, competing with AlphaFold2 on certain tasks
- Backed by Meta's substantial AI research infrastructure and $125-145 billion 2026 capex budget, ensuring continued computational resources for model development
Watch for
- Repository archived since August 2024, indicating Meta may have deprioritized active development of the ESM codebase in favor of newer AI initiatives like Meta Superintelligence Labs
- No dedicated commercial product or revenue stream; the project's long-term viability depends on Meta's continued willingness to fund open-source AI research without direct monetization
- Intense competition from Google DeepMind's AlphaFold3 and other protein prediction tools, which have larger dedicated teams and more frequent model updates
Recent moves
Key Information
- Industry
- AI Models & Architectures
- Founded
- 2020
Frequently Asked Questions
What is Meta AI ESM?
Meta AI ESM is an open-source deep learning project from Meta's FAIR team that applies large language model techniques to protein sequences. It predicts protein structure, function, and evolutionary relationships directly from amino acid sequences, helping researchers in computational biology.
Is the Meta AI ESM repository still active?
No, the ESM repository on GitHub was archived and set to read-only on August 1, 2024. This suggests Meta may have shifted research priorities toward newer AI initiatives, though the models remain widely used with over 4,100 stars and 794 forks.
How does Meta AI ESM compare to AlphaFold?
Meta AI ESM competes with Google DeepMind's AlphaFold in protein structure prediction. ESM has shown proven accuracy on benchmarks like CASP and CAMEO, rivaling AlphaFold2 on certain tasks. However, AlphaFold3 has a larger dedicated team and more frequent updates.
Can I use Meta AI ESM for drug discovery?
Yes, pharmaceutical and biotechnology companies use ESM for drug discovery, enzyme engineering, and synthetic biology. Its open-source MIT license allows broad adoption, and its predictions help identify protein targets and design novel molecules without direct revenue generation.
What is the license for Meta AI ESM?
Meta AI ESM is released under the permissive MIT license, enabling free use, modification, and distribution. This open-source approach has driven adoption across academia and industry, allowing researchers to integrate protein prediction into their workflows without licensing fees.
Why did Meta archive the ESM repository?
Meta archived the ESM repository in August 2024, likely due to shifting research priorities toward newer AI initiatives like Meta Superintelligence Labs. The company's $125-145 billion 2026 capex budget focuses on advanced AI systems, though ESM models remain available for use.
Sources
- github.com — Repository archived August 1, 2024; 4,100+ stars; 794 forks; MIT license; last commit June 2023
- investor.atmeta.com — Meta 2025 full-year revenue $200.97B; net income $60.46B; capex $72.22B; headcount context
- www.prnewswire.com — Meta Q1 2026 revenue $56.31B; EPS $10.44; headcount 77,986; capex forecast $125-145B; Meta Superintelligence Labs model release
- www.bloomberg.com — Meta in talks for Scale AI investment potentially exceeding $10 billion in June 2025