> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://developers.deepgram.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.deepgram.com/_mcp/server.

# August 28, 2025

> Deepgram Self-Hosted release 250828 adds a gpu_required health configuration option to fail fast at startup when no GPU is detected, for API 1.156.1 and Engine 3.100.0.

## Deepgram Self-Hosted August 2025 Release (250828)

### Container Images (release 250828)

* `quay.io/deepgram/self-hosted-api:release-250828`
  * Equivalent image to:
    * `quay.io/deepgram/self-hosted-api:1.156.1`
    * `quay.io/deepgram/onprem-api:release-250828`
    * `quay.io/deepgram/onprem-api:1.156.1`

* `quay.io/deepgram/self-hosted-engine:release-250828`
  * Equivalent image to:
    * `quay.io/deepgram/self-hosted-engine:3.100.0`
    * `quay.io/deepgram/onprem-engine:release-250828`
    * `quay.io/deepgram/onprem-engine:3.100.0`

  * Minimum required NVIDIA driver version: `>=570.172.08`

* `quay.io/deepgram/self-hosted-license-proxy:release-250828`
  * Equivalent image to:
    * `quay.io/deepgram/self-hosted-license-proxy:1.8.0`
    * `quay.io/deepgram/onprem-license-proxy:release-250828`
    * `quay.io/deepgram/onprem-license-proxy:1.8.0`

* `quay.io/deepgram/self-hosted-billing:release-250828`
  * Equivalent image to:
    * `quay.io/deepgram/self-hosted-billing:1.11.2`
    * `quay.io/deepgram/onprem-billing:release-250828`
    * `quay.io/deepgram/onprem-billing:1.11.2`

### This Release Contains The Following Changes

⚠️ **Please be advised that this release is known to contain an Aura-2 TTS defect where the wrong conditioning priors are applied, resulting in the insertion of a random utterance. While the issue is rare, we strongly advise all Aura-2 TTS customers to update to [release-250929](https://developers.deepgram.com/changelog/self-hosted-changelog#deepgram-self-hosted-september-2025-release-250929) as soon as possible, where this behavior was resolved by a code change.**

* **GPU Required Configuration** — New `[health]` section in engine.toml with `gpu_required` configuration option allows Engine to fail on startup if no GPU is detected. While Engine can run without a GPU, production deployments require one for acceptable performance. Set `gpu_required = true` to fail fast if no GPU is available, rather than running with severely degraded performance. Default: false.
* **General Improvements** — Keeps our software up-to-date.

**Important**: This is the last release that will include `onprem-*` image tags. The Deepgram image repositories have been updated to reflect our "self-hosted" naming. Images should now be pulled from the `self-hosted-*` Quay repositories. **Starting with the next release in September 2025, we will only publish new images to `self-hosted-*` repos, deprecating `onprem-*` repository variants.**