proxiML now supports Project Authorizers, letting you define an authorizer once per project and reuse it across every endpoint that needs it. Instead of re-entering the same API key or OIDC configuration in each endpoint job, create the authorizer a single time and reference it by ID when creating your endpoints.

proxiML now offers a dedicated Guardrail job type that lets you deploy inline safety and prompt guarding in front of any LLM endpoint. It currently provides three configurable protection layers — content safety, jailbreak detection, and injection detection — and exposes an OpenAI-compatible API so you can drop it in front of your existing service with no code changes.

proxiML now supports AMD accelerators as a compute backend for CloudBender nodes. Workloads can run on AMD GPUs through ROCm, expanding the hardware options available to CloudBender deployments beyond NVIDIA.

New job environments based on Python 3.14 are now available for supported frameworks.

proxiML now offers a dedicated LLM job type that lets you deploy pre-configured large language models as managed inference endpoints in minutes. Select a model family and size from the platform and get an OpenAI-compatible endpoint with no custom serving commands, Docker images, or manual checkpoint setup required. Currently supported families include Gemma 4, Qwen 3.5, and Qwen 3.6.

proxiML now supports Endpoint Authorizers, letting you control access to your deployed ML serving endpoints. Choose between API Key authentication for simple shared-secret access or OIDC (OpenID Connect) for enterprise-grade JWT-based authentication with configurable issuers, audiences, and required claims.

By default, uploaded artifacts use the <job_name>.zip convention (or <job_name>_<worker_number>.zip for multi-worker jobs), as described in the job form data section. You can now override those default artifact names (including per-worker names where applicable) from the job form and from the SDK via output_options.

The local connection capability has been redesigned for clearer workflows: connecting to jobs, datasets, checkpoints, and models that use Local storage or runtime access from your machine is now easier to discover and operate from the CLI and web UI.

CloudBender clean room stacks can now be managed through a clearer version upgrade workflow, so you can adopt platform fixes and capabilities in your CloudBender nodes on a controlled schedule.

New job environments based on Python 3.13 are now available for supported frameworks.