Getting started
Which model should I use?
Which model should I use?
See Choosing a model. H-Optimus for histology feature extraction, M-Optimus when you also need spatial gene-expression prediction, and tissue segmentation to filter background first.
What input size do the models expect?
What input size do the models expect?
224×224 tiles at 0.5 µm/px for embeddings and M-Optimus prediction; 512×512 at 8.0 µm/px for tissue segmentation.
Deployment & infrastructure
How is the model delivered?
How is the model delivered?
Two ways: an AWS SageMaker endpoint, or a containerized Docker image for on-premise. Both serve the same API. See Deployment.
What container runtime is required — OCI or Docker?
What container runtime is required — OCI or Docker?
The on-premise server ships as a Docker image loaded via
docker load and run with the NVIDIA Container Toolkit. Does it run on Kubernetes or another orchestrator?
Does it run on Kubernetes or another orchestrator?
TBD — the documented deployment is a single Docker container; confirm orchestration support.
What CPU architecture and GPU are required (on-prem)?
What CPU architecture and GPU are required (on-prem)?
x86-64 with an NVIDIA GPU of CUDA Compute Capability 8.6 (e.g. A10G-class) runs the container out of the box. Other architectures require recompilation. A 24 GB GPU (equivalent to SageMaker
ml.g5.xlarge) is a good baseline.What NVIDIA driver is needed, and is GPU access passthrough or shared?
What NVIDIA driver is needed, and is GPU access passthrough or shared?
Data security & privacy
Where does our data go? Does Bioptimus see it?
Where does our data go? Does Bioptimus see it?
Data stays entirely in your own environment (your AWS account, or your on-prem hardware). Bioptimus does not see your inputs, outputs, or usage. See Security & compliance.
What telemetry does Bioptimus collect?
What telemetry does Bioptimus collect?
Indicative (to be confirmed): disease area / task, slide volumes, functions called, and crash/bug signals — no slide images or patient data.
Model & SDK usage
Can I choose which embedding I get (CLS vs. supervised)?
Can I choose which embedding I get (CLS vs. supervised)?
No. H-Optimus returns the CLS token of size 1536; M-Optimus returns its MLP output. The embedding type is fixed.
What does an embedding look like?
What does an embedding look like?
A 1536-dimensional float vector per tile (H-Optimus and M-Optimus embedding mode). See the API reference.
What access methods are supported?
What access methods are supported?
Customer-managed SageMaker endpoints and on-premise containers; H-Optimus is also on Hugging Face for academic use.
How do I get the SDK?
How do I get the SDK?
Install the provided Python wheel (or via a package index once confirmed). See the SDK guide.
Commercial & support
What scientific support tiers are available?
What scientific support tiers are available?
Three tiers: Tier 1 — Co-Designed, Tier 2 — Supported, and Tier 3 — Self-Service.
Is there a free trial or proof-of-concept?
Is there a free trial or proof-of-concept?
TBD — contact Bioptimus to discuss evaluation options.
How do I avoid unexpected AWS charges?
How do I avoid unexpected AWS charges?
Delete SageMaker endpoints and batch models when finished — endpoints bill while running.