✳ ROOM TO EXPERIMENT. SPACE TO GROW.

A plan for your team. Compute for your workload.

Explore the proposed plans for individual developers and growing AI teams. Platform access and GPU usage are separate considerations; final prices and compute rates will be confirmed before paid workloads begin.

Free

Free

Explore the workflow

A place to learn the platform before committing compute.

Join the waitlist
  • Workspace access
  • Sample project walkthrough
  • Dataset preparation workflow
  • Community resources
Starter

Coming soon

Build your first custom model

For developers moving from experiments to a working model.

Request access
  • LoRA and QLoRA workflows
  • Training job monitoring
  • Model evaluation
  • Usage based GPU compute
Enterprise

Let’s talk

Fit your deployment needs

For teams with specific security and infrastructure requirements.

Contact the team
  • Everything in Pro
  • Dedicated compute options
  • Access control requirements
  • Deployment planning
Know the cost before you run.

GPU usage rates, endpoint charges, limits, and any minimum commitments will be provided before you activate paid workloads. The plans above are a proposed structure.

✳ FIND YOUR STARTING POINT

Compare the proposed plans.

A straightforward view of what each plan is designed for. Included limits and final entitlements are still being defined.

Platform capabilityFreeStarterProEnterprise
Who it’s forExploring the platformIndividual developersAI product teamsLarger deployments
Dataset workflowSample projectsOwn datasetsTeam datasetsRequirements based
Fine-tuningPreviewLoRA / QLoRALoRA / QLoRACompute options
EvaluationExample resultsRun comparisonTeam experimentsRequirements based
InferenceAPI walkthroughUsage basedEndpoint managementDedicated options
GPU computeNot includedUsage basedUsage basedAgreed separately

Proposed early access structure. No paid plan is activated by submitting a request.

✳ UNDERSTAND THE COMPUTE

What goes into model usage?

Your workload determines the resources you need. Review the rates and configuration before activating a paid job or endpoint.

Training compute

Model size, sequence length, GPU type, and run duration affect training usage. Parameter efficient methods can reduce memory requirements.

Data and artifacts

Datasets, saved checkpoints, and model versions use storage. Storage allowances and retention policies will be specified with each plan.

Inference serving

Endpoint configuration and request volume affect serving costs. Any reserved capacity, minimums, or usage rates will be shared before activation.

A LITTLE MORE CLARITY

Good questions.
Straight answers.

Have something else in mind?

Talk to the team
Is the Free plan a free GPU trial?

The proposed Free plan is for exploring the platform and sample workflows. It does not include a confirmed GPU allowance. Any compute credits or trial limits will be stated explicitly when access is offered.

How will GPU compute be charged?

Rates are still being finalized. You will receive the applicable GPU rates, serving charges, and any minimum commitments before a paid workload is activated.

Can I use my own datasets?

Starter and team plans are designed around your own datasets. Supported formats, size limits, retention, and available model versions will be confirmed during onboarding.

Can a team request dedicated infrastructure?

Enterprise requests can include dedicated compute, access control, and deployment requirements. Availability depends on the product scope and infrastructure options at the time.

Does requesting access create a paid subscription?

No. The request form expresses interest in early access. It does not create a paid account or authorize a charge.

Who owns the resulting model?

Model ownership and adapter export are core product goals. Rights remain subject to the foundation model’s license and the platform agreement provided before use.

YOUR NEXT CHAPTER IN AI

From your first dataset.
To your next release.

Bring a use case. Explore a model workflow built for your team.

Request access