Sign in
Pending benchmarksTracking since Nov 11, 2026

Is AlphaGenome 2 state of the art?Pending benchmarks.Announced Nov 11, 2026; no independent results yet.

Google DeepMind's next DNA sequence model, built to predict how DNA sequences and single-letter variants affect specific human cell types and cell states rather than broad tissue averages. Announced November 11, 2026. It is a conditional model: it takes a DNA sequence of up to 1 million letters plus one biological condition (a tissue, cell type, or state) at a time. The AlphaGenome API is available today for non-commercial research (Python client alphagenome.models.v2); commercial research is in private preview. No public weight release is stated. Intended for research use only, not clinical or diagnostic use.

Announced facts

The details so far

Lab
Google DeepMind
Announced
Nov 11, 2026
API
Live · alphagenome.models.v2
Open weights
Closed · No public AlphaGenome 2 weight release stated on the official blog, API docs, or the Google Hugging Face AlphaGenome collection, which lists AlphaGenome 1 fold repos only
Type
Conditional DNA sequence model that predicts molecular effects for one tissue, cell type, or state at a time, from a DNA sequence plus that condition. Scientists can also condition on their own single-cell RNA-seq snapshot via a cell encoder, without retraining (DeepMind blog)
Input length
Official docs list four AlphaGenome 2 input widths: 16,384 bp, 131,072 bp, 524,288 bp, and 1,048,576 bp (the 1-million-letter length also stated on the DeepMind blog; the docs recommend the longest width for accuracy)
Training data
Bulk and single-cell data from human and mouse samples, including over 107 million individual cell profiles across 68 tissues (DeepMind blog)
Outputs
The same core experimental readouts of DNA function as AlphaGenome, such as gene expression, histone modifications and chromatin accessibility, plus DNA methylation (DeepMind blog)
Hosted access
AlphaGenome API, free for non-commercial use, via a self-serve API key at the official AlphaGenome API page. Official docs identify the AlphaGenome 2 client as alphagenome.models.v2. Commercial research is in private preview; docs also point commercial and enterprise workflows to Google Cloud Model Garden on Gemini Enterprise Agent Platform (DeepMind blog, API page, and docs)
Eligibility
AlphaGenome Terms of Service: the API is for individuals and non-commercial organizations (universities, non-profits, research institutes, educational and government bodies), or journalism; it is not available to commercial entities on this API, even for non-commercial work. Credentials are personal and must not be shared (Terms of Service, last modified November 6, 2026)
Pricing
Official docs: the API is offered free of charge for non-commercial use, with query rates that vary with demand. The Terms of Service say Google does not currently charge for access but reserves the right to introduce fees with at least one month's notice
License
AlphaGenome Services Additional Terms of Service. API outputs and Atlas scores are for non-commercial research only, must not be used to train machine learning models, and must not be used for clinical decision-making. The Python client repository is licensed Apache 2.0; that license covers the client, not model weights
Clinical use
Intended for research use only; not validated or approved for clinical or diagnostic purposes (DeepMind blog and Terms of Service)
Related
Separate from AlphaGenome (also called AlphaGenome 1 in the official docs) and from AlphaGenome Atlas

Tags marked unsure are not stated on the lab's own pages.

Next steps

What would let us call it

Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.

  • Publicly announced with a source
  • Usable by the public (API or weights)
  • Listed on an independent board we track
  • Enough results for a rank in a category
  • Holds across two snapshots (crowns only)
Self-reported

What the lab claims

Self-reported. Shown for context; not used in any verdict.

Lab claim: self-reported identification of nearly twice as many variants that change gene expression in cells, compared with the previous model (DeepMind blog).

Lab claim: self-reported 30% vs 16% causal variants and 43% vs 23% directional gene effects on the TenK10K dataset of 28 primary immune cell types, compared with the original AlphaGenome fine-tuned on single-cell data (DeepMind blog).

Receipts

Sources

LABAlphaGenome 2: Deciphering how cells read DNA to better understand disease (Google DeepMind blog)deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/ · Nov 11, 2026LABAlphaGenome product page (Google DeepMind)deepmind.google/science/alphagenome/LABAlphaGenome APIdeepmind.google.com/science/alphagenome/apiLABAlphaGenome Services Additional Terms of Servicedeepmind.google.com/science/alphagenome/termsLABAlphaGenome documentationalphagenomedocs.com/LABAlphaGenome documentation: FAQsalphagenomedocs.com/faqs.htmlLABAlphaGenome documentation: Modelsalphagenomedocs.com/api/models.htmlREPOgoogle-deepmind/alphagenome on GitHubgithub.com/google-deepmind/alphagenomeLABGoogle DeepMind on X: AlphaGenome 2x.com/GoogleDeepMind/status/2120532824465805749 · Nov 11, 2026

Tracking continues until an independent board lists it.