Changes / AssemblyAI / Requests failing for Slam-1 model
AssemblyAI: Major incident: Requests failing for Slam-1 model
Summary
AssemblyAI reported a major partial outage affecting Slam-1 pre-recorded transcription requests, with slow processing, increased errors and failures. It started at 2025-10-22 17:56 UTC and was resolved at 2025-10-22 23:58 UTC, lasting 6 h 1 min. Universal requests were unaffected and offered as a workaround; Slam-1 performance returned to normal before resolution.
Machine-written from the diff below (Sol; type: incident outage). Check the diff before relying on it.
What changed
Before: not present
After: Requests failing for Slam-1 model
Before: not present
After: major
Before: not present
After: resolved
Before: not present
After: 2025-10-22 17:56 UTC
Before: not present
After: 2025-10-22 23:58 UTC
Before: not present
After: APIs - Asynchronous API, Asynchronous API
Before: not present
After: We are currently seeing longer-than-usual turnaround times for pre-recorded audio transcriptions that are using our Slam-1 model. Requests using the Universal model are unaffected. We are working to address this issue and will provide an update here as soon as possible.
Before: not present
After: We are now seeing increased error rates and failures for requests using the Slam-1 model so we are elevating this issue to a partial outage. Requests using Universal are working normally, so consider using that model as a short-term workaround if needed. Our team is working to resolve this issue as quickly as possible.
Before: not present
After: We are once again starting to see requests using Slam-1 succeed, but we are still working to fully resolve the issue. We will follow up with another update once the issue has been fully mitigated.
Before: not present
After: We are still seeing good success rates for requests using the Slam-1 model, but we are still seeing occasional slower-than-usual processing times, which we are working to address.
Before: not present
After: Success rates and turnaround times for requests using Slam-1 have returned to normal ranges, but we are continuing to monitor traffic to ensure continued good performance.
Before: not present
After: We have continued to see good performance for requests using Slam-1 and are now marking this issue as resolved.
First version of this incident archived by Rewindex. View the live incident (opens in new tab).
History of this incident
| When (UTC) | Change | Kind |
|---|---|---|
| Major incident: Requests failing for Slam-1 model (this change) AssemblyAI reported a major partial outage affecting Slam-1 pre-recorded transcription requests, with slow processing, increased errors and failures. It started at 2025-10-22 17:56 UTC and was resolved at 2025-10-22 23:58 UTC, lasting 6 h 1 min. Universal requests were unaffected and offered as a workaround; Slam-1 performance returned to normal before resolution. | Major incident |
All AssemblyAI changes → · AssemblyAI profile · Markdown · JSON