Sold by: Bioptimus
Deployed on AWS
Multimodal foundation model for oncology. M-Optimus-1 turns a routine H&E slide into ~6,000-gene spatial expression readouts. Pre-trained on millions of proprietary H&E whole-slide images (50+ tissues) plus paired patient cohorts with aligned H&E, bulk RNA-seq, and spatial transcriptomics.
Overview
Use M-Optimus-1 to build oncology models from H&E whole-slide images, optionally paired with bulk RNA-seq. Generate multimodal embeddings and reconstruct spatial gene expression across a curated panel of 6,000+ genes.
M-Optimus-1 is a multimodal foundation model for oncology, learning cross-modal representations across H&E, bulk RNA-seq, and spatial transcriptomics. Pre-trained using a unique proprietary dataset comprising millions of H&E whole slide images from more than 50 organ tissues, and thousands of patient records where H&E, bulk and spatial transcriptomics are paired and aligned.
Deploy M-Optimus-1 as an Amazon SageMaker model package inside your AWS account. Run real-time inference via SageMaker endpoints or batch inference on S3. Your data stays private, and is never accessed by Bioptimus.
Highlights
- M-Optimus-1 is a multimodal, multiscale foundation model for oncology that integrates information across H&E, bulk RNA-seq and spatial transcriptomics (10x Genomics Visium) to produce tile-, slide, and patient-level outputs.
- M-Optimus-1 can perform direct prediction from H&E (with optional bulk RNA-seq) to spatial expression across a curated pan-tissue panel of over 6,000 genes - tumor and immune cell biomarkers, cell-cell interactions, ligands and receptors, biological pathways, and drug candidates.
- M-Optimus-1 is multimodal by design, flexible in deployment. H&E-only inference still outperforms unimodal baselines when sequencing data is unavailable.
Details
Sold by
Delivery method
Latest version
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost |
|---|---|---|
ml.g5.xlarge Inference (Batch) Recommended | Model inference on the ml.g5.xlarge instance type, batch mode | $24,167.00/host/hour |
ml.g5.2xlarge Inference (Batch) | Model inference on the ml.g5.2xlarge instance type, batch mode | $24,167.00/host/hour |
inference.count.m.i.c Inference Pricing | inference.count.m.i.c Inference Pricing | $0.09/request |
Dimensions summary
You pay for M-Optimus-1 based on usage, with no upfront commitment. Two dimensions bill by host hours when you run model inference in batch mode. Each matches a different compute instance size: ml.g5.xlarge and ml.g5.2xlarge. The larger instance gives you more compute capacity per hour. A third dimension bills by inference request count, charging per request processed. Together, these let you choose between paying for the time your inference instance runs or paying per individual request. Your total cost scales with the instance you pick and how much inference you run.
Top-of-mind questions for buyers
It reads routine H&E slide images and predicts spatial gene expression across a panel of genes. It can run on the slide alone or combine the slide with bulk RNA-seq data to refine predictions. Each slide gets tiled, background removed, then processed to produce molecular readouts.
The two batch host-hour dimensions meter the time your chosen instance runs, so cost grows with runtime. The request dimension charges per inference request processed, independent of runtime. Host hours suit sustained batch jobs on many slides. Per-request suits lower, intermittent volumes where you pay only for what you submit.
The two batch host-hour dimensions meter the time your instance runs, so an active instance accrues charges whether or not it is actively processing. To stop host-hour charges, shut the instance down. The per-request dimension instead charges only per inference request processed, with no time-based accrual.
docs.bioptimus.com+1
Helpful?
Vendor refund policy
Non-refundable
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker model
An Amazon SageMaker model package is a pre-trained machine learning model ready to use without additional training. Use the model package to create a model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Version release notes
v1.3.0
- Combined predict-with-embedding: spatial gene expression and the tile embedding from a single M-Optimus forward pass.
- Slide-level aggregation: pool tile embeddings into a slide vector, or predicted expression into a bulk RNA-seq-like pseudobulk profile.
- Anonymized diagnostics endpoint (/diagnostics) and support bundles.
- Endpoint verification with typed error codes and remediation hints.
- Coordinate-driven tile extraction.
Reliability
- Empty/header-only bulk RNA input raises a clear typed error.
- Missing-gene sentinel preserved through bulk RNA preprocessing (no NaN embeddings).
- Slide-edge reads clamped; server reports terminal unhealthy state when all models fail to load; per-tile failures isolated.
Compatibility
- M-Optimus backbone, spatial gene-expression head, and tissue segmentation unchanged.
- Recommended instance: ml.g5.xlarge.
Additional details
Inputs
- Summary
H&E whole-slide images at 224×224 px tiles, 0.5 MPP.
Optional: H&E tiles at 512x512 px 8 MPP for tissue segmentation model. Paired bulk RNA-seq (.csv / .tsv / .txt / .h5ad). Exact Ensembl IDs match to the supported Ensembl ID list (see sample input data for details). All inputs should be base64 encoded json files. Inputs can contain multiple tiles at a time (jsonl format), however the total file size should not exceed 100MB for batch jobs and 6MB for real-time jobs.
- Limitations for input type
- Inputs must be in the specified json format with base64 encoded images. Please see sample input files for details.
- Input MIME type
- application/json
https://github.com/bioptimus/m-jumpstart/tree/main/data/input/real-time
https://github.com/bioptimus/m-jumpstart/tree/main/data/input/batch
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
image_data | Base64-encoded PNG of the tile | String | Yes |
slide_name | Identifier for the source slide | String | Yes |
x | Tile x-coordinate in the slide | Integer | Yes |
y | Tile y-coordinate in the slide | Integer | Yes |
width | Tile width in pixels | Integer | Yes |
height | Tile height in pixels | Integer | Yes |
patch_idx | Index of this tile | Integer | Yes |
model_name | Name of the model must be "m-optimus" | String | Yes |
mode | "prediction" (for gene expression) or "embedding" | String | Yes |
bulk_rna | Bulk RNA counts vector. If omitted, the server substitutes zeros. Must be in the exact gene order from the model's input_gene_set.csv. Pre-processed as log1p before passing to the model - so this should be raw counts. | List[float] | No |
Resources
Vendor resources
Support
Vendor support
Contact: support@bioptimus.com
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
The leading foundation model for computational pathology. H-optimus-1 turns H&E whole-slide images into task-agnostic embeddings for biomarker discovery, mutation prediction, and survival modeling, with state-of-the-art performance on 13 downstream tasks. 1.1B-parameter Vision Transformer pre-trained on 1M+ slides from 800,000+ patients across 50+ organ systems.
NOAA's Coastal Ocean Reanalysis (CORA) for the Gulf, East Coast/Atlantic, and Caribbean (GEC) is produced using verified hourly water levels from the National Ocean Service’s Center of Operational Oceanographic Products & Services (CO-OPS). ADvanced CIRCulation Model (ADCIRC) and Simulating WAves Nearshore (SWAN) models are coupled to model coastal water levels and nearshore waves. Hourly water level observations are used for data assimilation and validation to improve the accuracy of modeled water levels and wave datasets.
Additional Details:
Metadata associated with model domain and time span:
Timeseries - 1979 to 2022
Si[...]
mParticle is a real-time AI customer data platform that powers your entire marketing stack with high-quality customer data.
Announcements:
April 2, 2026: The Office of Water Prediction (OWP) has identified some brittle components of the AORC data processing pipelines that, by our estimations, have resulted in 0.04% of all available rows being incorrectly masked. We've resolved the issue and we are currently regenerating the Zarr files to ensure a complete record. We will post an updated announcement once the corrected data is posted.
The Analysis Of Record for Calibration (AORC) is a gridded record of near-surface weather conditions covering the continental United States and Alaska and their hydrologically contributing areas. It is defined on a latitude/longitude spatial grid with a mesh length of 30 arc seconds (~800 m), and a temporal resolution of one hour. Elements include hourly total precipitation, temperature, specific humidity, terrain-level pressure, downward longwave and shortwave radiation, and west-east and south-north wind components. It spans the period from 1979 acr[...]
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.