Deepgram Outage History

Deepgram's Voice Agent service experienced elevated error rates due to instability with Anthropic's Claude Sonnet 4.5 model, a third-party provider. The issue affected Voice Agent functionality while Deepgram's core infrastructure remained operational. The incident was resolved after approximately 7.5 days.

Deepgram's Voice Agent service experienced elevated error rates due to instability with Google Gemini LLMs, a third-party provider. Customers using bring-your-own Gemini LLMs encountered degraded performance and intermittent failures on LLM think requests, while Deepgram's core infrastructure remained healthy. The incident was resolved after approximately 10 days of monitoring and tracking the external provider's status.

Deepgram's Voice Agent API experienced elevated error rates due to issues with OpenAI models, affecting users who relied on a single LLM provider configuration. The service recommended configuring multiple LLM providers as a workaround to avoid downtime. The incident was resolved after 6.9 hours.

Deepgram experienced outbound networking issues that caused failed callbacks on STT PreRecorded API requests and TTS Batch requests, while also disrupting Voice Agent API sessions with third-party LLM and TTS providers. The incident affected the Batch API and TTS API components for 74.4 hours. A network infrastructure fix was implemented that immediately resolved the behavior, followed by continued monitoring until full resolution.

Deepgram's Voice Agent API experienced elevated error rates and latency issues with the NVIDIA Llama Nemotron Super 49B managed LLM model over a 76.5-hour period. The incident affected users relying on this specific LLM provider for voice agent functionality. Deepgram recommended configuring multiple LLM providers as a workaround to avoid downtime during the ongoing issues.

Deepgram's Voice Agent API experienced degraded service for 70 hours due to an ongoing OpenAI outage that affected the underlying LLM functionality. Users were advised to configure multiple LLM providers in their Voice Agent setup to avoid downtime during the incident. The service was fully restored after the OpenAI issues were resolved.

Deepgram's Voice Agent API experienced degraded performance for 8.9 hours due to an ongoing OpenAI outage that affected the underlying LLM service. The incident impacted Voice Agent functionality for users relying on OpenAI as their primary LLM provider. The issue was resolved after the OpenAI outage ended, with Deepgram recommending users configure multiple LLM providers to avoid similar downtime in the future.

An OpenAI outage caused issues with Deepgram's Voice Agent API for 26 minutes. Users were advised to configure multiple LLM providers in their Voice Agent setup to avoid downtime during the incident. The issue was resolved once the upstream OpenAI service was restored.

Deepgram's Streaming API experienced elevated errors and increased Speech-to-Text result latency for 8.9 hours. The issue affected the Streaming API component, causing delays in speech recognition processing. The incident was resolved after monitoring and investigation efforts.

Deepgram's Voice Agent API experienced issues due to an upstream Anthropic outage affecting Claude models, which began on February 17, 2026 at 20:15 UTC and lasted 20.5 hours. The incident impacted Voice Agent functionality for users relying on Anthropic's LLM services. The issue was resolved when the upstream Anthropic incident was fixed, restoring normal error rates and Voice Agent operations.