BioLM
Roadmap
Feedback
Changelog

Changelog

Follow up on the latest improvements and updates.

RSS

new

models

📚 New Model: ESM-1b

ESM-1b is now available on BioLM: https://biolm.ai/models/esm1b/
A widely adopted transformer protein LM for high-quality embeddings.
image

new

models

🧊 New Model: DeepViscosity

DeepViscosity is now available on BioLM: https://biolm.ai/models/deepviscosity/
Predict antibody viscosity profiles using paired heavy/light chains.
image

new

models

🧠 New Model: DSM-150M Base

DSM-150M Base is now available on BioLM: https://biolm.ai/models/dsm-150m-base/
A compact, efficient protein embedding model suitable for fast inference.
image

new

models

🧠 New Model: DSM-650M Base

DSM-650M Base is now available on BioLM: https://biolm.ai/models/dsm-650m-base/
A high-capacity embedding model for deeper protein sequence representation.
image

new

models

🔗 New Model: DSM-650M PPI

DSM-650M PPI is now available on BioLM: https://biolm.ai/models/dsm-650m-ppi/
Use this model to estimate interaction likelihood between two protein sequences.
image

new

models

⚡ New Models: E1-150M, E1-300M, and E1-600M

The E1 model family is now available on BioLM:
• E1-150M - https://biolm.ai/models/e1-150m/
• E1-300M - https://biolm.ai/models/e1-300m/
• E1-600M - https://biolm.ai/models/e1-600m/
These transformer-based protein language models provide fast, high-quality sequence embeddings across three capacity levels, enabling tailored performance for different throughput and accuracy needs.
image

new

models

🔷 New Model Release: DIAMOND

We’ve added
DIAMOND
, a fast and efficient protein structure search model built for large-scale discovery workflows. DIAMOND enables high-throughput structural comparisons with strong accuracy, making it well-suited for iterative design loops and database-scale similarity search.
Highlights
  • Ultra-fast structural search at scale
  • Designed for proteome- and library-level comparisons
  • Initial database includes USPTO, with more to come!
  • Use it to screen designed molecules to make sure they're not patented already.
Check it out here: https://biolm.ai/models/diamond/
image

new

models

🧠 New Model Releases

We've expanded our model collection with three additions focused on protein stability and molecular interaction modeling.
🕒 TemBERTure GitHub →
Predicts
protein thermostability class
and
melting temperature (Tm)
using sequence data from thermophilic and non-thermophilic organisms.
🌍 SPURS GitHub →
Predicts
ΔΔG stability changes
from mutations and helps identify
functionally important protein sites
.
⚡ Boltz-2 GitHub →
Predicts
3D protein–ligand structures
and
binding affinities
, enabling fast, large-scale molecular screening.
---
All three models are now available
to use within the platform.

new

console

protocols

🧬 Introducing Protocols: Reproducible AI Workflows for Molecular Design

Protocols
are reproducible, end-to-end workflows that define how BioLM models and tools work together to achieve a design goal. Think of them as
experiment blueprints
: structured, versioned, and repeatable.
What’s New
  • 🧪
    Protocols Tab:
    A new home in Console for running and managing your workflows.
  • 💡
    Built-In Protocols:
    Start with
    AntiFold Design Antibodies
    , an end-to-end pipeline for generating and scoring antibody variants.
  • ⚙️
    Custom Protocols:
    Define your own protocol logic using a simple YAML schema — including input parameters, batch limits, and Pub/Sub configuration.
  • 📊
    Interactive Results:
    Each run produces a detailed dashboard showing folding stages, RMSD calculations, ranked variant scores, etc.
  • 🔁
    Versioned & Shareable:
    Duplicate, modify, and rerun protocols using GitHub for consistent, reproducible design studies.
Why It Matters
Protocols make it easy to
codify best practices
in your molecular design workflow — reducing manual setup, improving traceability, and helping teams standardize complex AI-driven experiments.
Try It Now
Go to
Console → Protocols
to explore the built-in
AntiFold Design Antibodies
workflow or upload your own YAML blueprint to create a new one.
Run a Protocol or learn more in our help docs.

new

models

🆕 New Models Released: TEMPRO-3B & TEMPRO-650M

🚀 New models are live on BioLM!
We’ve just released TEMPRO-3B and TEMPRO-650M models.
  • TEMPRO-3B
    : ~3B params, high accuracy for complex tasks
  • TEMPRO-650M
    : ~650M params, lightweight & fast
Both are available now via the API and Playground. Give them a spin and let us know your feedback! 🎉
Load More
→