Hosted tool
Validate a protein sequence in your browser, see what an ESMFold2 result looks like with the interactive example below, and fold your own single-chain protein up to 700 amino acids on our hosted GPUs.
Paste a sequence or FASTA to validate it instantly. Running the fold is hosted on our GPUs.
Hosted runs use usage-based credits and include biosafety screening. Do not submit pathogen, toxin, regulated, confidential, or unauthorized third-party sequences.
Precomputed Ubiquitin MSA output from the Biohub ESMFold2 tutorial.
Live example: Ubiquitin (76 residues, pTM 0.814) folded with ESMFold2. Open the full example result.
Paste a raw sequence or upload a FASTA file. The tool parses it in the browser, strips FASTA headers and whitespace, and checks the length and residue alphabet so you know it is valid before spending anything.
When you run a fold, the sequence is sent to our self-hosted ESMFold2-Fast weights, screened for biosafety, and returned as an mmCIF structure with pTM and pLDDT confidence you can inspect in the same 3D viewer shown here.
A downloadable predicted structure, per-structure confidence metrics, and an interactive viewer — without installing PyTorch, provisioning a GPU, or cloning the repository.
Prefer to run it yourself? Generate browser-local starter code with the ESMFold2 code generator, or wire it into your own pipeline via the ESMFold2 API.
Validating a sequence and exploring the interactive example structure are free in the browser. Running a hosted ESMFold2 fold on your own sequence needs a free account and usage-based credits, because each run uses real GPU time and biosafety screening.
Single protein chains of 16 to 700 standard amino acids (codes ACDEFGHIKLMNPQRSTVWY). Multi-chain complexes, RNA/DNA, ligands, and MSA workflows are not part of this hosted beta.
No. Sequence validation runs entirely in your browser. When you submit a hosted run, the sequence is processed transiently to fold the structure and is not retained as a stored sequence afterwards.
Each result reports per-structure pTM and pLDDT confidence. See the ESMFold2 benchmark for measured runtime and confidence across sequence lengths before relying on a prediction.
Written by ESMFold2 Tools Editorial Team
Fact-checked against Biohub official docs, GitHub examples, and linked primary sources.
Last updated June 7, 2026