Model type
Multitrack generative protein transformer
ESM3 generates and completes protein sequence, structure and functional annotations using a shared multimodal representation.
Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.
Multitrack generative protein transformer
Complete or partial protein sequence, structure and function-keyword tracks.
Completed sequence, structure and functional tracks.
Official project documentation and implementation: https://github.com/evolutionaryscale/esm
limited source coverage · Automated source review, 2026-09-16. All specifications and missing details
12 evaluations · 12 results. Different protocols are not a single leaderboard.
Applied filters: All linked evaluations
| Tested configuration | Protocol and dataset | Finding | Evidence and details |
|---|---|---|---|
| Configuration: ESM3 | Task: PFMBench ANTI-RES: Antibiotic resistance Dataset subset: Antibiotic resistance (PFMBench split) | 0.584 accuracy fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench ANTI-RES: Antibiotic resistance Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Anti.Res.) |
| Configuration: ESM3 | Task: PFMBench BINDING-DB: BindingDB Dataset subset: BindingDB (PFMBench split) | 0.225 spearman correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench BINDING-DB: BindingDB Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Bind. DB) |
| Configuration: ESM3 | Task: PFMBench CLONING-CLF: Cloning CLF Dataset subset: Cloning CLF (PFMBench split) | 0.774 auroc fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench CLONING-CLF: Cloning CLF Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Clo. CLF) |
| Configuration: ESM3 | Task: PFMBench DEEPLOC2: DeepLoc2 Multi Dataset subset: DeepLoc2 Multi (PFMBench split) | 0.659 f1 fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench DEEPLOC2: DeepLoc2 Multi Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(DL2 M.) |
| Configuration: ESM3 | Task: PFMBench DEEPSOL: DeepSol Dataset subset: DeepSol (PFMBench split) | 0.781 auroc fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench DEEPSOL: DeepSol Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(DeepSol) |
| Configuration: ESM3 | Task: PFMBench EC: Enzyme Commission Dataset subset: Enzyme Commission (PFMBench split) | 0.648 f1 fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench EC: Enzyme Commission Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(EC) |
| Configuration: ESM3 | Task: PFMBench MAT-PROD: Material production Dataset subset: Material production (PFMBench split) | 0.775 accuracy fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench MAT-PROD: Material production Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Mat.Pro.) |
| Configuration: ESM3 | Task: PFMBench METAL-ION: Metal ion binding Dataset subset: Metal ion binding (PFMBench split) | 0.703 accuracy fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench METAL-ION: Metal ion binding Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(M. I. Bin.) |
| Configuration: ESM3 | Task: PFMBench PDB-BIND: PDBbind Dataset subset: PDBbind (PFMBench split) | 0.156 spearman correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench PDB-BIND: PDBbind Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(PDBBind) |
| Configuration: ESM3 | Task: PFMBench PROTEINGYM-ZS: ProteinGym zero-shot variant effect prediction Dataset subset: ProteinGym (PFMBench split) | 0.414 spearman correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench PROTEINGYM-ZS: ProteinGym zero-shot variant effect prediction Scored zero-shot, with no fine-tuning, as reported in PFMBench Table 4. Aggregation: Not reported pfmbench primary benchmark evidence · Table 4, row(ESM3 [ 18 ]), column(ProteinGym) |
| Configuration: ESM3 | Task: PFMBench SEC-STRUCT: Secondary structure Dataset subset: Secondary structure (PFMBench split) | 0.813 accuracy fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench SEC-STRUCT: Secondary structure Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Sec. Str.) |
| Configuration: ESM3 | Task: PFMBench STABILITY: TAPE_Stability Dataset subset: TAPE_Stability (PFMBench split) | 0.157 spearman correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM3 on PFMBench STABILITY: TAPE_Stability Fine-tuned with an adapter under the PFMBench harness; train, validation and test counts are in Table 1. Aggregation: Not reported pfmbench primary benchmark evidence · Table 3, row(ESM3 [ 18 ]), column(Stability) |
Source checking is not independent reproduction. Release 2026-09-29-06401fd5b220.
Related profile: ESM3. This page retains the exact record and its evaluation context.
Protein foundation model evaluated by the PFMBench authors under their fine-tuning protocol. Input family: Sequence-Structure-Function.
ESM3 generates and completes protein sequence, structure and functional annotations using a shared multimodal representation. Transformer generative masked-language model with discrete sequence, structure and function tracks; generation iteratively fills masked positions. The documented inputs are complete or partial protein sequence, structure and function-keyword tracks. The output consists of completed sequence, structure and functional tracks.
Published March 2024 small/medium/large models, August 2024 small/medium models, and esm3-sm-open-v1 local weights. The applicable input limits require configuration-specific checking.
Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied.
Stable record: discovery-model-esm3Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.
| Property | Description and evidence |
|---|---|
| Model type | Multitrack generative protein transformerSources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Architecture | Transformer generative masked-language model with discrete sequence, structure and function tracks; generation iteratively fills masked positions.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Inputs | Complete or partial protein sequence, structure and function-keyword tracks.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Outputs | Completed sequence, structure and functional tracks.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Parameters | 1.4B small, 7B medium and 98B large variants.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Known versions | Published March 2024 small/medium/large models, August 2024 small/medium models, and esm3-sm-open-v1 local weights.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Training data | ESM3 overview reports 2.78 billion proteins and 771 billion unique tokens for the largest model; this is not a verified per-checkpoint manifest.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Training cutoff | The inspected ESM3 release documentation does not supply one latest-data date shared across its sequence, structure and function tracks. · Not reported in inspected sourcesSources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Context limits | The inspected ESM3 family documentation does not establish one context limit for local small, hosted medium and hosted large models; a specific service/checkpoint is required. · Not reported in inspected sourcesSources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Weights licence | MIT stated in the inspected ESM3 README; model and API access conditions remain separate.Sources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Access | Official project documentation and implementation: https://github.com/evolutionaryscale/esmSources (3)evolutionaryscale/esm: README.md; evolutionaryscale/esm: _assets/ESM3_README.md; evolutionaryscale/esm: THIRD_PARTY_NOTICE.md · _assets/ESM3_README.md: overview, ESM3 Family, Running Locally and Licenses |
| Code licence | MITSourcesevolutionaryscale/esm: LICENSE.md · LICENSE.md: licence text |
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
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-esm3 Individual claims | pfmbench primary benchmark evidence Section 3.3 Supported Models; Table 2 model identity and Table 3 named row Version: 2506.14796v1 | source checked automated source review · 2026-09-23 Audit detailsSource review establishes this relationship only. Exact evaluated configurations and original numerical review status remain unchanged. Source identifies ESM-2 650M, ESM-C 600M and ESM3 as the respective foundation models evaluated with adapter tuning. Retain per-task configurations and metrics. Field: Claim: model-evaluation-identity-436d088fde334db49648 Source artifact SHA-256: Hash scope: Exact retrieved primary paper artifact bytes. |
View linked audit checks and correction history
Release 2026-09-29-06401fd5b220 · Record review: source checked
Stable ID: pfmbench-method-esm3