EvolutionaryScale

EvolutionaryScale is an American artificial intelligence company focused on biology, founded in February 2024 by Alex Rives, Salvatore Candido, Tom Sercu, and other former members of the Meta AI / FAIR ESM (Evolutionary Scale Modeling) team that had developed the ESM, ESM-2, and ESMFold protein language models. The company is headquartered in New York with an additional San Francisco office and develops the ESM3 protein language model (released June 2024 in 1.4B, 7B, and 98B variants) and an ongoing program for protein-design and biology-AI applications. As of April 2026, EvolutionaryScale has backing from Lux Capital, Nat Friedman and Daniel Gross's NFDG, Amazon, NVIDIA, and other investors, with a Series A round of $142 million at announcement, and is positioned alongside Isomorphic Labs, Inceptive, and other biology-AI peers in the 2024 to 2026 commercial protein-AI market.

At a glance

• Founded: February 2024 in New York City and San Francisco, by Alex Rives, Salvatore Candido, Tom Sercu, and other former Meta AI / FAIR ESM team members.
• Status: Private. Stealth founding through early 2024; public launch June 2024 with the ESM3 release.
• Funding: $142 million Series A at announcement (June 2024). Backers include Lux Capital, Nat Friedman and Daniel Gross's NFDG, Amazon, NVIDIA, and other investors.
• CEO: Alex Rives, Co-Founder and Chief Scientist. Machine-learning researcher; founding lead of the Meta AI / FAIR ESM team.
• Other notable leadership: Salvatore Candido, Co-Founder and Chief Technology Officer. Tom Sercu, Co-Founder.
• Open weights: Yes, partial. ESM3 1.4B and 7B variants are released open-weights through Hugging Face under noncommercial-research licensing; the 98B variant is available commercially through the EvolutionaryScale Forge platform.
• Flagship products: ESM3 (protein language model, June 2024, in 1.4B, 7B, and 98B variants), EvolutionaryScale Forge (the commercial platform for ESM3 and other protein-AI applications), and continued protein-language-model research.

Origins

EvolutionaryScale was founded in February 2024 by Alex Rives, Salvatore Candido, Tom Sercu, and other former members of the Meta AI / FAIR ESM team. The founders had built the ESM, ESM-2, and ESMFold protein language models at Meta from 2019 through 2023, with ESMFold (released November 2022) demonstrating that protein-language-model approaches could match or exceed AlphaFold-style structural-prediction accuracy on protein-prediction benchmarks while running faster than equivalent AlphaFold inference.

The 2023 transition saw Meta restructure the ESM team during the broader Meta AI restructuring period that also affected adjacent research groups. The ESM team's continued collective interest in protein-language-model research, combined with a commercial opportunity in protein design and biology-AI applications, anchored the founding of EvolutionaryScale in February 2024.

The June 2024 ESM3 release was the company's first public-facing milestone. ESM3 (1.4B, 7B, and 98B variants) extended the protein-language-model research lineage with the multi-track architecture (jointly modeling protein sequence, structure, and function) and scale increases beyond ESM-2. The ESM3 1.4B and 7B variants were released open-weights through Hugging Face under noncommercial-research licensing; the 98B variant is available commercially through the EvolutionaryScale Forge platform.

The 2024 to 2026 period has seen continued ESM3 iteration and the Forge platform commercial expansion. Industry coverage has consistently characterized EvolutionaryScale as one of the principal protein-language-model labs globally, alongside Isomorphic Labs (the Alphabet subsidiary developing AlphaFold-derived drug-discovery applications) and Inceptive (the RNA-foundation-model startup founded by former Google researchers).

Mission and strategy

EvolutionaryScale's stated mission is to use AI to understand biology and accelerate scientific discovery. The strategic premise is that protein language models, trained on the corpus of known protein sequences (UniProt and other databases), can capture the evolutionary patterns governing protein structure and function, and that scaling these models in the manner of large language models will enable protein-design capability at commercial scale.

The strategy combines three threads. First, the ESM3 protein-language-model research with continued model iteration through 2024 to 2026. Second, the EvolutionaryScale Forge commercial platform providing commercial access to the larger ESM3 variants and other protein-AI applications. Third, published research output extending the ESM and ESMFold research lineage at major academic venues.

The competitive premise is that the ESM team's continuity (with Alex Rives, Salvatore Candido, Tom Sercu, and other senior researchers transitioning together from Meta AI / FAIR), combined with published research output and open-weights release of smaller variants, gives EvolutionaryScale a structural advantage in commercial protein-AI applications. Industry coverage has consistently characterized the founding-team continuity as a strategic advantage compared to peer protein-AI labs without the ESM research lineage.

Distribution channels include the EvolutionaryScale Forge commercial platform for the larger ESM3 variants, open-weights distribution through Hugging Face for the smaller variants, published research output through major academic venues, and enterprise and research-customer relationships across pharmaceutical, biotech, and other industries.

Models and products

• ESM3. Protein language model released June 2024 in 1.4B, 7B, and 98B variants. Multi-track architecture jointly modeling protein sequence, structure, and function. Smaller variants open-weights through Hugging Face under noncommercial-research licensing; the 98B variant available commercially through Forge.
• EvolutionaryScale Forge. Commercial platform providing commercial access to the larger ESM3 variants and other protein-AI applications. Direct API and developer-platform access.
• Continued protein-language-model research. Published research output through 2024 to 2026 extending the ESM and ESMFold research lineage.

Distribution channels include the Forge commercial platform, open-weights distribution through Hugging Face, published research output through major academic venues, and enterprise and research-customer relationships.

Benchmarks and standing

EvolutionaryScale's evaluation framework focuses on protein-structure-prediction benchmarks (CASP-style protein-structure assessment), protein-design benchmarks (zero-shot protein-design capability), and published research output. ESM3 has been characterized in protein-AI industry coverage as one of the principal protein-language-model families globally, with multi-track capability that distinguishes it from peer protein-AI lab outputs.

The ESM and ESMFold research lineage from the Meta AI / FAIR period (2019 to 2023) had established the protein-language-model research direction, with ESMFold's 2022 demonstration that protein-language-model approaches could match AlphaFold-style structural prediction accuracy. EvolutionaryScale's continued ESM3 iteration extends that research lineage with multi-track architecture and scale increases.

The Forge commercial platform's 2024 to 2026 commercial expansion has continued the enterprise-customer base across pharmaceutical, biotech, and other industries. Direct comparison benchmarks against Isomorphic Labs' AlphaFold-derived commercial offerings and other peer commercial platforms remain limited.

Leadership

As of April 2026, EvolutionaryScale's senior leadership includes:

• Alex Rives, Co-Founder and Chief Scientist. Machine-learning researcher; founding lead of the Meta AI / FAIR ESM team.
• Salvatore Candido, Co-Founder and Chief Technology Officer. Machine-learning engineer; senior member of the Meta AI / FAIR ESM team.
• Tom Sercu, Co-Founder.
• Senior research-and-engineering leadership across the ESM3 research, Forge commercial-platform, and other application-area programs.

Continued senior research recruitment has supported the 2024 to 2026 ESM3 product transition and the Forge commercial-service expansion.

Funding and backers

• Series A (June 2024): $142 million at announcement, with Lux Capital, Nat Friedman and Daniel Gross's NFDG, Amazon, NVIDIA, and other investors.
• Earlier capital: Stealth-period seed and angel capital before the June 2024 public launch.

The backing from Lux Capital (with deep-tech and biology-AI investing thesis), NFDG (the Friedman-Gross investing partnership with AI-startup positioning), Amazon (the AWS strategic-customer relationship), and NVIDIA (the AI compute strategic-investor relationship) provides EvolutionaryScale with financial-runway certainty alongside strategic-investor cooperation.

Industry position

EvolutionaryScale occupies a structurally distinctive position as one of the principal commercial protein-language-model labs globally, with ESM research lineage continuity from the Meta AI / FAIR ESM team, the ESM3 protein-language-model release, the Forge commercial platform, and backing from Lux Capital, NFDG, Amazon, and NVIDIA. Industry coverage has consistently characterized EvolutionaryScale as one of the principal protein-AI labs globally, alongside Isomorphic Labs and Inceptive in the 2024 to 2026 commercial protein-AI market.

The 2024 to 2026 commercial protein-AI market has seen enterprise-customer adoption across pharmaceutical, biotech, materials-discovery, and other industries. EvolutionaryScale's open-weights release of the smaller ESM3 variants (under noncommercial-research licensing) has anchored the company's open-research positioning alongside commercial Forge platform expansion.

Competitive landscape

• Isomorphic Labs. Direct competitor on commercial protein-AI applications. Isomorphic's AlphaFold-derived commercial offerings represent the principal alternative to ESM3 in commercial drug-discovery and biology-AI applications.
• Inceptive. Biology-AI peer with different RNA-foundation-model focus.
• Meta AI / FAIR ESM legacy. Founding-team origin organization. Meta has continued AI research output but has shifted resources away from the protein-AI research focus.
• Google DeepMind AlphaFold and Isomorphic Labs. Precursor research lineage and the AlphaFold research community.
• Recursion, Insitro, Cradle Bio, Generate Biomedicines. Biology-AI peers with different drug-discovery and protein-design focus.
• Hippocratic AI. Healthcare-AI peer with different clinical-application focus.
• Hugging Face. Open-weights distribution partner; EvolutionaryScale distributes the smaller ESM3 variants through Hugging Face.

Outlook

• The continued cadence of ESM3 iteration and next-generation protein-language-model release timeline through 2026 to 2027.
• The Forge commercial platform expansion and the continued enterprise-customer base growth across pharmaceutical, biotech, and other industries.
• The continued ESM and ESMFold research lineage extension through published research output.
• The competitive dynamic with Isomorphic Labs on commercial protein-AI applications.
• The Series B or adjacent fundraising timeline through 2026 to 2027.
• Continued senior research-talent recruitment and senior leadership stability through commercial expansion.

Sources

• EvolutionaryScale official site. Company reference.
• EvolutionaryScale Forge. Commercial platform reference.
• ESM3 release announcement. June 2024 ESM3 announcement.
• EvolutionaryScale on Hugging Face. Open-weights model distribution.
• ESM team research history. Predecessor Meta AI / FAIR ESM repository.