Literature · ZhejiangLab
Genos 1.2B
A 1.2-billion-parameter genomic foundation model for high-efficiency sequence analysis and prediction tasks.
What Genos 1.2B is
Genos 1.2B is a genomic foundation model developed by ZhejiangLab. Built with 1.2 billion parameters, it is designed for high-efficiency genome sequence analysis and supports single-nucleotide resolution learning alongside million-base-pair context modelling.
The model forms part of the broader Genos suite of foundation models for the human genome, serving as a general-purpose base model spanning multiple organisms. It provides a computational foundation that researchers can adapt for downstream sequence-level and gene-level prediction tasks.
How it works
Genos 1.2B operates as a foundation model capable of processing genomic sequences with million-base-pair context lengths and single-nucleotide resolution. Its architecture is associated with Mixtral-based structure and is also implemented within the Megatron-LM framework for training and deployment.
Key facts
- Parameter count
- 1.2 billion
- Context length
- Million-base-pair context modelling
- Architecture tag
- mixtral
- Minimum GPU memory
- 24 GB
- Framework variants
- Hugging Face safetensors, Megatron-LM
What you can use it for
- Downstream Sequence-Level Prediction
- Serves as a pretrained base model for fine-tuning on sequence-level functional and structural prediction tasks across genomes.
- Gene-Level Prediction
- Provides sequence representations to drive downstream prediction tasks targeted at individual genes.
- Single-Nucleotide Resolution Analysis
- Evaluates genomic sequences at single-nucleotide resolution across extended million-base-pair sequence contexts.
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
Genos 1.2B
Strengths and limits
Good at
- Achieves million-base-pair context modelling for genomic sequences.
- Provides single-nucleotide resolution learning capability.
- Designed for high-efficiency genome sequence analysis at a 1.2-billion-parameter scale.
- Spans multiple organisms as a base model for diverse downstream applications.
Watch out for
- Requires a minimum of 24 GB of GPU memory to run.
- Represents the smallest variant in the Genos family, offering lower capacity compared to the 10-billion-parameter models.
Inputs and outputs
You provide
Prompt
The model continues from here.
You get back
The model's completion.
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 Genos 1.2B 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 Genos 1.2B 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 organisms does Genos 1.2B support?
While developed primarily as part of a foundation model family focused on the human genome, it functions as a general-purpose model spanning multiple organisms.
What is the context window capacity of Genos 1.2B?
The Genos model collection achieves million-base-pair context modelling while maintaining single-nucleotide resolution.
Are there other model sizes available in the Genos series?
Yes, alongside Genos 1.2B, the collection includes Genos-10B and Genos-10B-v2, both offering larger parameter scales for complex genomic tasks.
Which frameworks support Genos 1.2B weights?
The weights are available as standard safetensors checkpoints as well as checkpoints trained under the Megatron-LM framework.
Written from Genos 1.2B'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.