Model type
Biomolecular structure predictor family
Configuration
Boltz-1 (MSA)
Boltz predicts biomolecular complex structures; Boltz-2 also predicts binding affinity.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 LimitationsHow it worksBoltz workflow
Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 LimitationsOverview
Inputs
Protein, nucleic-acid and ligand specifications in prediction input files.
Outputs
Predicted complex structures and, for supported Boltz-2 inputs, binding-affinity predictions.
Access
Official project documentation and implementation: https://github.com/jwohlwend/boltz
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitationslimited source coverage · Automated source review, 2026-09-23. All specifications and missing details
Evaluations and results
1 evaluation · 1 result. Different protocols are not a single leaderboard.
Filter evaluations
Applied filters: All linked evaluations
| Tested configuration | Protocol and dataset | Finding | Evidence and details |
|---|---|---|---|
| Configuration: Boltz-1 (MSA) | Protocol: ESMFold2 Runs N’ Poses reported comparison msa: Runs N’ Poses ligand pass rate (MSA) Dataset subset: Runs N’ Poses complete-case intersection: 2,573 scored ligands (ESMFold2 Runs N’ Poses reported comparison split) | 60% ligand_pass_rate percent · higher Uncertainty: Not reported Coverage: unit: ligands; scored: 2573; eligible: unreported; note: Complete-case intersection; source 2600 systems is not a ligand denominator. | Author-reported evaluation · Source checkedMethods, coverage and sourceRuns N’ Poses receptor–ligand co-folding; source benchmark 2,600 systems. Figure 2C reports n=2,573 scored ligands on the intersection where all models produced valid predictions, after excluding undefined SuCOS scores. Multiple ligands in one system are scored independently. Five seeds × five diffusion samples per target; select top candidate by ipTM. Success requires lDDT-PLI >0.8 and BiSyRMSD <2 angstrom. Baselines use 10 recycles and 200 diffusion steps; ESMFold2 uses 10 or 20 loops as labelled and truncated 68-step diffusion. Single-sequence and MSA conditions remain separate. Aggregation: Not reported ESMFold2 primary paper v1, Figure 2C Runs N’ Poses · PDF page 5, Figure 2C, Runs N’ Poses subpanel (right), msa block, bar 7 from left (Boltz-1 (MSA)), exact printed bar label |
Source checking is not independent reproduction. Release 2026-09-29-06401fd5b220.
Use this model
How it works, versions and access
Related profile: Boltz. This page retains the exact record and its evaluation context.
This configuration
Author-evaluated folding configuration with conditioning and loop count retained from Figure 2C.
- record
- Boltz-1 (MSA)
- configuration
- Not reported
- entity type
- Configuration
How it works
How it works
Boltz-2 first encodes the molecular inputs, alignments and optional templates into token and pair features. A Pairformer trunk updates those features and conditions atom-coordinate diffusion to generate a complex. Separate confidence and affinity modules assess the prediction; affinity classification and regression outputs answer different questions. This describes the Boltz-2 generation; a Boltz-1 result must retain its original checkpoint and prediction procedure.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 LimitationsVersions and reproducibility
Boltz-1 and Boltz-2 are distinct released generations; the catalogue does not select an evaluated checkpoint. The Boltz-2 report describes training crops up to 768 tokens. This is a training-crop size rather than a universal inference maximum; affinity additionally uses a pocket crop.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 LimitationsStrengths, limitations and unresolved questions
Strengths and limitations
Strengths and considerations
- The project distributes prediction code, model weights and training instructions.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations
Limitations and conditions
- Affinity predictions depend on a plausible binding pose and mix biochemical endpoint types. The report notes limited handling of cofactors, water and multimeric binding partners, and substantial variation between assays.
Sources (6)
jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations
Profile review details
Follow-up review of Training data, Training cutoff, Context limits, Parameters, Code licence, Weights licence, Known versions. Source locations and before/after decisions are recorded in the 23 September profile-evidence audit. Other explanatory content retains its earlier source scope. No human scientific review or independent reproduction is implied.
Stable record: discovery-model-boltzSpecifications
Inputs, training, access and other details
Explanatory profile: limited source coverage · Automated source review, 2026-09-23. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.
| Property | Description and evidence |
|---|---|
| Model type | Biomolecular structure predictor familySources (6)jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations |
| Architecture | Boltz-1 and Boltz-2 are separate generations. In the inspected Boltz-2 implementation, molecular/MSA/template embeddings enter a Pairformer trunk, which conditions atom-coordinate diffusion; confidence and affinity are separate output modules.Sources (6)jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations |
| Inputs | Protein, nucleic-acid and ligand specifications in prediction input files.Sources (6)jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations |
| Outputs | Predicted complex structures and, for supported Boltz-2 inputs, binding-affinity predictions.Sources (6)jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations |
| Parameters | A complete parameter count tied to the released structure, confidence and affinity checkpoints is not stated in the inspected report or model constructor. These components must be identified before comparing parameter totals. No total was inferred from file size or architecture names. · Not reported in inspected sourcesSources (2)Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction; jwohlwend/boltz: src/boltz/model/models/boltz2.py · XML sectionsS5-S12 Architecture and Training; src/boltz/model/models/boltz2.py Boltz2.__init__ component construction. |
| Known versions | Boltz-1 and Boltz-2 are separate model generations. In the Boltz-2 report, Boltz-2x means Boltz-2 with physicality steering potentials enabled; it is an inference condition, not evidence of a separately trained checkpoint.Sources (2)Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction; jwohlwend/boltz: README.md · README Introduction; XML sectionS15 paragraphP32 (Boltz-2x definition and Boltz-1 comparison). |
| Training data | Boltz-2 structure training uses experimental PDB structures released before 2023-06-01, MISATO/ATLAS/mdCATH molecular-dynamics ensembles, AlphaFold2 monomer predictions and Boltz-1 predictions of several complex types. Affinity regression uses filtered PubChem, ChEMBL and BindingDB assays; binary supervision uses PubChem HTS, CeMM and MIDAS plus synthetic decoys.SourcesBoltz-2: Towards Accurate and Efficient Binding Affinity Prediction · XML sectionS3 paragraphsP14-P16 Structural Data; sectionS4 paragraphsP20-P21 Binding Affinity Data. |
| Training cutoff | 2023-06-01 applies to release dates of experimental PDB structures used by Boltz-2. The source does not give a common cutoff for molecular dynamics, distilled predictions and affinity databases; this date must not be assigned to Boltz-1 or to all Boltz-2 training evidence.SourcesBoltz-2: Towards Accurate and Efficient Binding Affinity Prediction · XML P14 experimental PDB cutoff; P16 distillation and P20-P21 affinity sources. |
| Context limits | Boltz-2 training crops extend to 768 molecular tokens. This is a training setting, not a universal inference limit. Affinity training also crops predicted binding pockets; permitted input size depends on the actual inference implementation and resources.SourcesBoltz-2: Towards Accurate and Efficient Binding Affinity Prediction · XML sectionS6 paragraphP23 Trunk optimization; sectionS12 paragraphP29 Affinity training. |
| Weights licence | The official README explicitly releases all code and weights under MIT for academic and commercial use. This establishes distribution terms, not the identity of a checkpoint used by an external paper.Sourcesjwohlwend/boltz: README.md · README.md Introduction line19. |
| Access | Official project documentation and implementation: https://github.com/jwohlwend/boltzSources (6)jwohlwend/boltz: README.md; jwohlwend/boltz: docs/prediction.md; jwohlwend/boltz: docs/training.md; jwohlwend/boltz: src/boltz/model/models/boltz2.py; jwohlwend/boltz: scripts/train/configs/full.yaml; boltz2: Journal full-text XML · src/boltz/model/models/boltz2.py: Boltz2.__init__, forward, PairformerModule, AtomDiffusion and AffinityModule; docs/prediction.md: affinity outputs; Boltz-2 report Sections 2 Data, 3 Architecture, 4 Training and 6 Limitations |
| Code licence | MIT for code at commit b1ebfc46ecf57f5414e0d1a6f9027bbb122c53bc.Sources (2)jwohlwend/boltz: LICENSE; jwohlwend/boltz: README.md · LICENSE MIT text; README.md Introduction line19. |
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
1 evidence row matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| Relationship: family discovery-model-boltz Individual claims | ESMFold2 primary paper v1, Figure 2C Runs N’ Poses Figure 2C Runs N’ Poses, bar label Boltz-1 (MSA); Appendix A.2.10 Version: 10.64898/2026.06.03.729735v1; posted 2026-06-04 | source checked automated source review · 2026-09-23 Audit detailsSource-backed evaluated identity only; no independent reproduction. Field: Claim: esmfold2-2026-runs-n-poses-method-boltz-1-msa-discovery-model-boltz-identity-claim Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Sources and history
View linked audit checks and correction history
Release 2026-09-29-06401fd5b220 · Record review: source checked
1 source records and release history
- ESMFold2 primary paper v1, Figure 2C Runs N’ Poses · Original source · 10.64898/2026.06.03.729735v1; posted 2026-06-04
Technical metadata and extraction receipts
Stable ID: esmfold2-2026-runs-n-poses-method-boltz-1-msa
- areas
- proteins-complexes
- source locator
- Figure 2C Runs N’ Poses, bar label Boltz-1 (MSA); Appendix A.2.10
- missing metadata
- checkpoint revision: unreported; parameters: unextracted