Deepgram self-hosted 261001 drops Whisper: what Sume STT callers do
Deepgram Self-Hosted release 261001 removes Whisper support and asks you to move to Nova-3 first. What it means for Sume STT, which has no model to migrate.
Deepgram's October 1, 2026 Self-Hosted release 261001 says the Engine no longer supports Whisper models, and that deployments using Whisper should migrate to Nova-3 before upgrading. That only affects people running Deepgram's containers. A caller of Sume STT sends audio_url to /v1/stt-1.0/transcribe and never names a provider model, so there is nothing of that kind to migrate.
Sources: the Deepgram changelog and Sume's API reference, read 2026-10-01.
What does release 261001 change?
| Item | What the page says |
|---|---|
| Whisper | Engine no longer supports Whisper models; migrate to Nova-3 before upgrading |
| Flux TTS watermarking | Action required: add the watermarker model file and set watermarker_uuid; Engine will not start with Flux TTS enabled if it is missing |
| FIPS images | Metrics use the engine_ prefix and /v1/certificates is served |
| Formatting | Numerals support Japanese (language=ja); spoken letters kept in entities such as W123 |
Who needs to act?
Teams that run the self-hosted Deepgram stack and still send Whisper requests. The page gives the order: move to Nova-3 first, then upgrade. Teams that call a hosted API are not described as affected by this item.
What stays the same for a Sume STT caller?
The Sume route description says provider model ids stay internal and callers use the public id sume/stt-1.0, and the request schema says provider knobs such as diarize and tag_audio_events are fixed server-side. A caller sets audio_url, optionally language_code, and optionally duration_seconds (1 to 600) to improve the usage reservation.
What is the trade-off?
You do not run or upgrade anything, but you also cannot choose or hold an engine version. If your compliance or latency needs require a specific self-hosted stack, this route does not replace it. For a hosted-versus-hosted view of retirements, read the Whisper shutdown post.
Sources
Related posts
More in Developers
- Deno fetch stops retrying fresh-connection POSTs: Sume keys
Deno now retries fetch transport errors only on reused connections. That cuts duplicate POSTs, but a paid Sume submit still needs an Idempotency-Key.
- Remove video speckle noise by API: median filter radius
Sume's video filter allowlists median, which replaces each pixel with the middle value of its neighbours. Radius runs 1 to 127; start at 1 and compare.
- Extract audio from MP4 to WAV or MP3: the detach defaults
Sume audio detach returns WAV (pcm_s16le, sample-exact) by default or MP3 at 128 kbps. Pick WAV for speech-to-text and timeline use, MP3 when file size matters.
- Extract audio segment start beyond video length: error and fix
Audio detach fails detach_start_past_source when range.start is past the probed duration. It differs from audio_detach_range_empty. Check duration first.
Written by Sume