What is ModelRun / Modular?
The former ModelRun route now redirects to Modular, which offers managed and self-hosted AI inference.
Core offerings
- Shared serverless endpoints billed per token.
- Dedicated endpoints billed by provisioned GPU time.
- Custom-model deployment in Modular Cloud, a customer cloud or self-hosted infrastructure.
- A free Community Edition of MAX and Mojo for self-hosted use.
Pricing
As of 1 October 2026, Modular describes shared endpoints as per-token, dedicated endpoints as per-minute or GPU-hour, and custom deployments as quote-based. It does not expose a complete public numeric model-rate table. Buyers should use Modular's live Cloud console or contact sales for an exact quote.
API and model availability
Because no complete public numeric model catalogue is published, ModelRun / Modular is not given speculative API Cost rows.
Official sources: Modular platform, Modular pricing.