Chai Discovery
Open-source-first structure-prediction company whose Chai-2 model claims near-20% hit rates on de novo antibody design.
1. Core Product / Service
Chai Discovery builds foundation models that predict and reprogram interactions between biochemical molecules. Chai-1 (released 2024) is an open-source foundation model for molecular structure prediction that performs at state-of-the-art level [1]. Chai-2 extends to fully de novo antibody design: given a target antigen and epitope, it designs all complementarity-determining regions (CDRs) from scratch, reporting a 16–20% hit rate across 52 novel antigen targets while testing only ~20 designs per target — versus ~0.1% for prior computational methods [1][2].
2. Target Users & Pain Points
- Biotech/pharma discovery teams who want a permissively-licensed structure/antibody model instead of a closed commercial platform.
- The open-source scientific community — Chai-1's open weights position it as the community alternative to alphafold3's restricted release.
Pain solved: antibody discovery has historically screened millions-to-billions of candidates; Chai-2's claimed hit rate collapses the search space, and open weights lower the adoption cost.
3. Competitive Landscape
| Player | Model | Vs. Chai |
|---|---|---|
| alphafold3 | Unified structure prediction (Diffusion) | AF3 is closed/restricted; Chai-1 is open-source |
| evolutionary-scale | ESM3 sequence/structure/function | ESM3 is licensed to pharma; Chai-1 is open |
| absci | IgDesign1 + wet lab | Absci is integrated + clinical; Chai is model-first |
| nabla-bio | JAM antibody design | Nabla is partnership-funded; Chai is open-source-first |
4. Unique Observations
- Open-source as a wedge: Chai-1's open weights let it capture the research community's mindshare the way alphafold3 did not (AF3's weights remain gated) — a deliberate "open model → commercial platform" funnel [1].
- The near-20% claim is a benchmark, not a product: Chai-2's reported hit rate (16–20% testing only 20 designs) is a headline-grabbing number, but it is an internal benchmark that independent teams have yet to reproduce at scale — treat as directionally strong, not settled (per verify-not-trust).
- Ex-Stripe + ex-OpenAI lineage: co-founders Jack Dent (ex-Stripe) and Josh Meier (ex-Absci/OpenAI/Meta) bring product-engineering discipline to a field dominated by academic labs [1].
5. Financials / Funding
- Seed: $30M (led by Thrive Capital, OpenAI, Dimension) [2].
- Series A: $70M (Aug 6, 2025; led by Menlo Ventures / Anthology Fund, with DST Global, Yosemite, SV Angel, DCVC, existing investors) — total $100M [1][2].
- Board: Mikael Dolsten (former Pfizer CSO) joined the board [1].
6. People & Relationships
- Co-founders: Joshua Meier (CEO; ex-Absci, Facebook AI, OpenAI), Jack Dent (ex-Stripe), Matthew McPartlon, Jacques Boitreaud [1].
- Investors: Thrive Capital, OpenAI, Menlo Ventures (Anthology Fund), DST Global, Dimension.
- Peers: alphafold3, evolutionary-scale, absci, nabla-bio.
Sources
- [1] Business Wire, "Chai Discovery Announces $70 million Series A To Transform Molecular Design" (2026-08-24)
- [2] FinSMEs, "Chai Discovery Raises $70M in Series A Funding" (2026-08-24)