Multi-Modal Foundation · mist-models
MIST 26.9M · melting point
Predicts molecular melting points in degrees Celsius from SMILES representations using a 26.9M-parameter encoder.
What MIST 26.9M · melting point is
MIST 26.9M · melting point is a molecular property prediction model developed by the Electrochemical Energy Group at the University of Michigan, Ann Arbor. It adapts a foundation encoder pretrained on molecular structures to perform supervised regression of melting-point values directly from chemical notations.
Melting point is a key physical property determining a substance's thermal operating window. By predicting melting temperatures directly from chemical structure without requiring physical synthesis, the model assists researchers in identifying viable molecular candidates across chemistry, materials science, and energy storage.
How it works
The model is based on the RoBERTa-PreLayerNorm architecture featuring 8 layers, 8 attention heads, a hidden dimension of 512, an intermediate size of 2,048, and a maximum sequence context of 2,048 tokens. The base encoder was pretrained on Kekulized SMILES strings from the Enamine REAL Space database using a masked language modelling objective with cross-entropy loss.
For melting-point regression, the encoder is coupled to a task network consisting of a two-layer multi-layer perceptron with Gaussian Error Linear Unit (GELU) activations and dropout. Input representations are pooled from the first token's hidden state. Supervised fine-tuning was performed using mean squared error loss on 5,734 melting-point measurements curated from PubChem.
Key facts
- Architecture
- RoBERTa-PreLayerNorm (8 layers, 8 heads) with a 2-layer MLP regression head
- Parameter count
- 26.9M
- Pretraining dataset
- Enamine REAL Space
- Fine-tuning dataset
- PubChem Characteristic Temperatures dataset (5,734 entries)
- Context length
- 2,048 tokens
- Input format
- Kekulized SMILES strings
- Output
- Melting point in degrees Celsius (°C) or mean-pooled feature vectors
What you can use it for
- Thermal Window Assessment
- Screening candidate organic molecules to determine their melting temperatures and operational thermal boundaries.
- Electrolyte Formulation
- Evaluating prospective battery electrolyte constituents to ensure they remain liquid at designated operating temperatures.
- Molecular Representation Extraction
- Generating dense vector embeddings from SMILES strings for downstream cheminformatics and machine learning tasks.
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 · melting point
Strengths and limits
Good at
- Fine-tuned on experimentally verified melting points filtered for small molecules with fewer than 20 heavy atoms and zero radical electrons.
- Pretrained on Enamine REAL Space, anchoring structural representations within synthetically accessible chemical space.
- Compact 26.9M-parameter footprint enables rapid property estimation on standard computational hardware.
Watch out for
- Restricted to small organic compounds containing fewer than 20 non-hydrogen heavy atoms.
- Not calibrated for radical species, having been trained strictly on molecules with zero radical electrons.
- Requires SMILES inputs to be Kekulized and relies on the SMIRK tokenizer.
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 · melting 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 · melting 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 molecular inputs does the model accept?
The model accepts Kekulized SMILES strings representing chemical structures, processed via the SMIRK tokenizer.
In what units are predictions returned?
The regression head outputs melting-point predictions in degrees Celsius (°C).
What is the maximum molecular size supported?
The model was fine-tuned on molecules filtered to fewer than 20 heavy atoms, excluding hydrogen.
Can this model be used purely for feature extraction?
Yes, the model can output mean-pooled vector representations of input molecular sequences instead of temperature values.
How to cite it
Wadell, A., Bhutani, A., Azumah, V., Ellis-Mohr, A. R., Kelly, C., Zhao, H., Nayak, A. K., Hegazy, K., Brace, A., Lin, H., Emani, M., Vishwanath, V., Gering, K., Alkan, M., Gibbs, T., Wells, J., Varshney, L. R., Ramsundar, B., Duraisamy, K., Mahoney, M. W., Ramanathan, A., & Viswanathan, V. (2025). Foundation Models for Discovery and Exploration in Chemical Space. arXiv:2510.18900.
Written from MIST 26.9M · melting 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.