Multi-Modal Foundation · mist-models
MIST 26.9M · boiling point
Predicts molecular normal boiling points in degrees Celsius directly from SMILES strings.
What MIST 26.9M · boiling point is
MIST 26.9M · boiling point is a molecular property prediction model developed by the Electrochemical Energy Group at the University of Michigan, Ann Arbor. It is part of the Molecular Insight SMILES Transformers family and is designed to calculate normal boiling points from simplified molecular-input line-entry system (SMILES) chemical strings.
A chemical compound's boiling point defines a crucial part of its thermal operating window, governing volatility and phase transitions. By providing computational predictions from chemical representations, the model assists researchers in assessing the physical properties of organic molecules without preliminary synthesis or physical laboratory measurement.
How it works
The model architecture uses a RoBERTa-PreLayerNorm encoder featuring 8 layers, 8 attention heads, a hidden size of 512, an intermediate size of 2048, and a maximum sequence context of 2048 tokens. The base encoder was pretrained on molecular SMILES from the Enamine REAL Space database using a masked language modelling objective with cross-entropy loss and the FusedLAMB optimizer.
For property regression, the pretrained encoder is attached to a task network comprising a two-layer multilayer perceptron with Gaussian Error Linear Unit (GELU) activations and dropout. Hidden state vectors are pooled by taking the representation of the sequence's first token. Fine-tuning was executed using the AdamW optimizer and mean squared error loss on 3,969 boiling-point records curated from PubChem, restricted to compounds with fewer than 20 non-hydrogen heavy atoms and zero radical electrons.
Key facts
- Architecture
- RoBERTa-PreLayerNorm encoder (8 layers, 8 heads, hidden size 512) with a two-layer MLP task head
- Base model
- mist-28M-ti624ev1
- Context length
- 2048 tokens
- Pretraining data
- Enamine REAL Space
- Fine-tuning data
- 3,969 PubChem records from the Characteristic Temperatures dataset
- Input format
- Kekulized SMILES strings
- Output format
- Normal boiling point in degrees Celsius (°C)
What you can use it for
- Thermal Operating Window Estimation
- Estimate the boiling points of candidate organic compounds to determine liquid-phase thermal boundaries.
- Electrolyte and Solvent Screening
- Screen prospective liquid battery electrolyte candidates and solvents for volatility constraints.
- Virtual Library Filtering
- Filter large compound databases by boiling temperature thresholds before conducting synthesis.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
MIST 26.9M · boiling point
Strengths and limits
Good at
- Pretrained on the Enamine REAL Space dataset covering synthetically accessible chemical space.
- Utilises a 2048-token context window capable of accommodating long molecular strings.
- Employs a PreLayerNorm transformer architecture combined with a lightweight MLP task network.
Watch out for
- Training and fine-tuning were restricted to molecules containing fewer than 20 non-hydrogen heavy atoms.
- Cannot reliably handle free radicals as the fine-tuning dataset excluded open-shell radical compounds.
- Expects Kekulized SMILES inputs and depends on the SMIRK tokenizer, which requires a Rust environment.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use MIST 26.9M · boiling point commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run MIST 26.9M · boiling point without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on CPU only, typically in 12s, and a run costs about 28 credits. Visitors without an account get 1,000 credits a month.
What input structure does the model require?
The model expects molecular SMILES strings, specifically Kekulized SMILES, processed using the SmirkTokenizerFast tokenizer.
In what unit are boiling points reported?
Predictions are direct scalar regression values representing the normal boiling point in degrees Celsius (°C).
What evaluation metrics were used during testing?
The model was evaluated using Root Mean Squared Error (RMSE) across a random 10% test split of the curated dataset.
Can this model predict boiling points for large polymers or macromolecules?
No, the model was fine-tuned exclusively on small organic molecules with fewer than 20 heavy atoms.
How to cite it
Wadell, Alexius, Anoushka Bhutani, Victor Azumah, Austin R. Ellis-Mohr, Celia Kelly, Hancheng Zhao, Anuj K. Nayak, et al. 'Foundation Models for Discovery and Exploration in Chemical Space.' arXiv preprint arXiv:2510.18900 (2025). https://doi.org/10.48550/arXiv.2510.18900
Written from MIST 26.9M · boiling point's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.