Multi-Modal Foundation · mist-models
MIST 1.8B · Cv
Predicts constant-volume heat capacity from molecular SMILES strings for thermodynamic and chemical screening.
What MIST 1.8B · Cv is
MIST 1.8B · Cv is a molecular property prediction model created by the Electrochemical Energy Group at the University of Michigan, Ann Arbor. It predicts the heat capacity at constant volume at 298.15 K (Cv) directly from molecular SMILES (Simplified Molecular Input Line Entry System) strings. Heat capacity indicates how much thermal energy is needed to raise the temperature of a molecule, which is fundamental to chemical thermodynamics and energy storage modelling.
The model belongs to the Molecular Insight SMILES Transformers (MIST) family of chemical foundation models. It uses an encoder pre-trained on large synthetic chemical spaces, coupled to a regression head to estimate quantum chemical properties without requiring explicit ab initio quantum chemistry calculations.
How it works
The core encoder is based on the RoBERTa-PreLayerNorm architecture, featuring 24 layers, a hidden dimension of 2,304, an intermediate size of 9,216, and 18 attention heads, supporting inputs up to 2,048 tokens. It was initially pre-trained on SMILES strings from the Enamine REAL Space dataset using a masked language modelling objective and cross-entropy loss.
For downstream prediction, the representation is pooled by taking the final hidden state of the first token in the sequence. This embedding is passed to a task network comprising a two-layer multilayer perceptron with Gaussian Error Linear Unit (GELU) activations and dropout. The network was fine-tuned on QM9 dataset heat capacity values using mean squared error loss.
Key facts
- Architecture
- RoBERTa-PreLayerNorm encoder with a 2-layer MLP task head
- Parameters
- 1.8 billion
- Pre-training data
- Enamine REAL Space
- Fine-tuning data
- QM9 Heat Capacity at 298.15K (MoleculeNet)
- Context length
- 2,048 tokens
What you can use it for
- Heat Capacity Prediction
- Estimating molecular heat capacity at constant volume at 298.15 K (Cv) in cal/(mol·K) from chemical SMILES.
- Thermodynamic Screening
- Evaluating candidate molecules such as electrolytes, drug compounds, and synthetic intermediates for thermal and thermodynamic properties.
- Molecular Feature Extraction
- Extracting dense latent representations and mean-pooled embedding vectors from SMILES inputs for cheminformatics pipelines.
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 1.8B · Cv
Strengths and limits
Good at
- Pre-trained on the Enamine REAL Space library, capturing structural relationships across synthetically accessible chemical space.
- Accommodates sequences up to 2,048 tokens, enabling evaluation of extended molecular SMILES strings.
- Utilises first-token pooled representations with a non-linear GELU multi-layer perceptron tailored for continuous property regression.
Watch out for
- Restricted to predicting heat capacity (Cv) at a single reference temperature (298.15 K).
- Fine-tuning is derived from the QM9 benchmark, which covers small organic molecules.
- Tokenisation relies on SMIRK tokenisers, which require an active Rust installation.
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 1.8B · Cv 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 1.8B · Cv 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 1× A100 16GB, typically in 12s, and a run costs about 70 credits. Visitors without an account get 1,000 credits a month.
What input formats does the model accept?
The model takes molecular SMILES strings as inputs, specifically expecting Kekulized SMILES representation.
What units are used for the predicted heat capacity?
The regression head outputs heat capacity at constant volume (Cv) at 298.15 K in cal/(mol·K).
How are molecular representations pooled for regression?
The model extracts the final hidden state vector of the first token in the sequence to represent the entire molecule before feeding it into the task network.
Can the model output vector embeddings?
Yes, it can function as a feature-extraction model to output pooled hidden state vectors per input sequence.
How to cite it
Alexius Wadell, Anoushka Bhutani, Victor Azumah, Austin R. Ellis-Mohr, Celia Kelly, Hancheng Zhao, Anuj K. Nayak, Kareem Hegazy, Alexander Brace, Hongyi Lin, Murali Emani, Venkatram Vishwanath, Kevin Gering, Melisa Alkan, Tom Gibbs, Jack Wells, Lav R. Varshney, Bharath Ramsundar, Karthik Duraisamy, Michael W. Mahoney, Arvind Ramanathan, Venkatasubramanian Viswanathan. Foundation Models for Discovery and Exploration in Chemical Space. arXiv:2510.18900 [physics], 2025. https://doi.org/10.48550/arXiv.2510.18900
Written from MIST 1.8B · Cv'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.