
TxGemma
TxGemma is a collection of open AI models built on Google DeepMind's Gemma that leverages large language models to improve therapeutic development efficiency by predicting properties of therapeutic entities throughout the drug discovery process.
https://developers.googleblog.com/en/introducing-txgemma-open-models-improving-therapeutics-development?ref=aipure

Product Information
Updated:May 9, 2025
TxGemma Monthly Traffic Trends
TxGemma experienced a 13.6% growth in traffic, reaching 1.8M visits. This moderate growth could be attributed to the recent updates and announcements from Google, including the Agent Development Kit (ADK) and the Agent2Agent (A2A) protocol, which may have increased interest and engagement among developers.
What is TxGemma
TxGemma is Google's latest healthcare-focused AI initiative and the open successor to Tx-LLM, designed specifically for drug development and therapeutic research. Fine-tuned from Gemma 2 using 7 million training examples, TxGemma comes in three sizes (2B, 9B, and 27B parameters) and offers both prediction and conversational capabilities. The models are trained to understand and analyze various therapeutic modalities and targets, including small molecules, proteins, nucleic acids, diseases, and cell lines.
Key Features of TxGemma
TxGemma is a collection of open AI models built on Google DeepMind's Gemma platform, specifically designed to improve therapeutic drug development efficiency. Fine-tuned with 7 million training examples, it comes in three sizes (2B, 9B, and 27B) and offers both prediction and conversational capabilities for analyzing therapeutic data, understanding molecular structures, and predicting drug properties throughout the discovery process.
Multiple Model Sizes: Available in three parameter sizes (2B, 9B, and 27B), each offering different levels of capability and computational requirements for various therapeutic tasks
Dual Functionality Modes: Includes both 'predict' versions for specific therapeutic tasks and 'chat' versions for conversational analysis and explanation of reasoning
Fine-tuning Capability: Supports custom fine-tuning on proprietary data, allowing researchers to adapt the model to their specific therapeutic research needs
Integration with Agentic-Tx: Can be integrated into a larger system with 18 specialized tools for complex research problems, including PubMed search and molecular analysis tools
Use Cases of TxGemma
Drug Property Prediction: Predicting crucial properties of potential therapeutic compounds, such as toxicity levels and blood-brain barrier penetration
Clinical Trial Outcome Analysis: Helping researchers predict and analyze potential outcomes of clinical trials, reducing the risk of failure in later stages
Molecular Structure Analysis: Analyzing and understanding complex molecular structures and their interactions for drug development
Research Workflow Optimization: Streamlining the drug discovery process by providing AI-driven insights and recommendations for research directions
Pros
Open-source nature allows for community contribution and improvement
Strong performance compared to specialized single-task models
Flexible architecture supporting both prediction and conversation modes
Cons
Chat versions show slightly reduced performance compared to prediction-only versions
Requires significant computational resources for larger models
Still limited by the general challenges of AI in drug discovery
How to Use TxGemma
Access TxGemma: Access TxGemma through either Vertex AI Model Garden or Hugging Face platforms where the models are available
Choose Model Size: Select from three available sizes (2B, 9B, or 27B parameters) based on your needs. Each size has a 'predict' version for specific tasks and 'chat' versions (9B and 27B only) for conversational interactions
Basic Inference: Use the inference Colab notebook provided by Google to run basic predictions on therapeutic tasks like molecule toxicity, classification, or regression tasks
Fine-tune for Custom Tasks: Use the fine-tuning Colab notebook to adapt TxGemma to your specific therapeutic data and tasks. This allows customization for proprietary research needs
Implement Conversational Features: For the chat versions (9B and 27B), utilize the model's ability to explain reasoning and engage in multi-turn discussions about therapeutic predictions
Advanced Usage with Agentic-Tx: Use the Agentic-Tx Colab notebook to implement more complex workflows combining TxGemma with other tools like PubMed search, molecular tools, and gene/protein tools
Validate Results: Test the model's predictions against known benchmarks and validate results for your specific use case
Share Feedback: Contribute to the improvement of the open model by sharing feedback with the development community
TxGemma FAQs
TxGemma is a collection of open AI models developed by Google, fine-tuned from Gemma 2, designed to improve the efficiency of therapeutic development. It's specifically trained to understand and predict the properties of therapeutic entities throughout the drug discovery process.
TxGemma Video
Popular Articles

Gemini 2.5 Pro Preview 05-06 Update
May 8, 2025

Suno AI v4.5: The Ultimate AI Music Generator Upgrade in 2025
May 6, 2025

How to Install and Use FramePack: The Best Free Open-Source AI Video Generator for Long Videos in 2025
Apr 28, 2025

DeepAgent Review 2025: The God-Tier AI Agent that's going viral everywhere
Apr 27, 2025
Analytics of TxGemma Website
TxGemma Traffic & Rankings
1.8M
Monthly Visits
-
Global Rank
-
Category Rank
Traffic Trends: Feb 2025-Apr 2025
TxGemma User Insights
00:00:33
Avg. Visit Duration
1.51
Pages Per Visit
73.42%
User Bounce Rate
Top Regions of TxGemma
US: 25.51%
IN: 10.44%
VN: 5.24%
KR: 5.2%
JP: 3.25%
Others: 50.36%