Project AURIONAI Use-Cases Research & Innovation
Boltz PBC
scientific

Boltz-2

Co-folding and binding-affinity model; vendor claims it approaches FEP accuracy at about 1000x speed; evaluated on FEP+, CASP16, MF-PCBA.

Boltz PBCBoltz PBCUSA / UKMIT (Jameel Clinic)MIT (Jameel Clinic)USARecursionRecursion
Bar chart of Boltz-2 binding affinity Pearson correlation against other methods.
Levelmodel
Classscientific
Parametersnot in the notes
Contextnot in the notes
ReleasedJun 2025
Opennessopen source
Runs onsingle datacentre GPU
Statusactive

Benchmark Results

vendor-reported
FEP+/OpenFE binding affinity benchmarkdomain
About 0.62; about 1,000x faster than FEPvendor-reported
0.62
00.350.7

Score in Pearson r. Higher is better.

Jun 2025 intuitionlabs.ai, github.com
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

successor
Boltz-1 to Boltz-2: successor, high confidenceBoltz-1Boltz-1Boltz PBC, Nov 2024Boltz-2Boltz-2Boltz PBC, Jun 2025
Boltz PBCBoltz-1 is the parent (successor)high confidencenotes
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.

Industry Use Cases

Healthcare
Protein structure and complexes

R&D Topics

Biomolecular structure prediction and co-foldingexemplar

Identity

Developer
Boltz PBC, MIT (Jameel Clinic), Recursion
Family
not in the notes
Released
Jun 2025 sourcessecondary report
Weights published
–
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
diffusion AF3-style co-folding + binding affinity head
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Context length
not in the notes
Input
molecule
Output
molecule
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open source
Licence
MIT License sources
Access routes
open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
single datacentre GPU
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Sources

  1. Open source protein models: OpenFold vs Boltz scalingbiotech.substack.com https://scalingbiotech.substack.com/p/open-source-protein-models-openfold
  2. GitHub - jwohlwend/boltz: Official repository for the Boltz biomolecular interaction models github.com https://github.com/jwohlwend/boltz
  3. intuitionlabs.ai/articles/open-source-ai-drug-discovery-boltz-... https://intuitionlabs.ai/articles/open-source-ai-drug-discovery-boltz-pfizer
  4. boltz.com/boltz2 https://boltz.com/boltz2
  5. par.nsf.gov/servlets/purl/10585253 https://par.nsf.gov/servlets/purl/10585253

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Boltz-2", https://aurion.semai.id/models/boltz-boltz-2/, records of October 2026.