Jul Patch 1

v3.1-40 Update 1

17 Jul, 2026

This update release primarily includes bug fixes and minor enhancements to existing features listed below.

Token Factory

Display URL by Modality

The API endpoint section in the model card now displays endpoint URLs dynamically based on the model’s supported modalities and API compatibility. If a model supports multiple endpoints, all applicable URLs are displayed to the user.

In the first example, the model supports chat completions, responses and messages APIs.

In the next example, the model supports APIs for image generation.

---

Storage Settings for Huggingface Cache Volume

Operators can now specify storage settings such as "read/write mode" and "storage class" for Huggingface cache volumes. This allows the storage volume to be used as a common model cache across multiple model deployments allowing these deployments to reuse already-downloaded model weights instead of downloading the same weights repeatedly.

In the example below, for the model deployment based on vLLM, the operator can now specify the "storage class name" and "read write mode"

---