Is Reflection AI Down?

Issues are mostly recovered. --- [2026-11-11T01:32:26Z] (monitoring) **Recovering:** From about 00:30 to 01:17 UTC, most non-streaming requests to Beam-1.0-501B-A23B failed with 502 errors after 20–60 seconds, and streaming requests ran at reduced throughput. This was caused by a network issue between our API gateway and one of our inference providers. Since 01:17 UTC, throughput and success rates have largely recovered, and we're continuing to monitor recovery. --- [2026-11-11T01:14:11Z] (identified) We are having a major API availability issue

The API has been operating normally since 21:38 UTC. We’re working with the provider to prevent a recurrence, and we apologize for the disruption. --- [2026-11-03T21:43:20Z] (monitoring) From about 20:10 to 21:38 UTC, many requests to Beam-501B-A23B failed with errors or rate limits, or were much slower than usual, because of a network problem at one of our inference providers. Since 21:38 UTC, error rates and response times have been back to normal. We’re watching closely and working with the provider on the root cause. Other models were not affected. --- [2026-11-03T21:42:40Z] (investigating) **Recovering:** Between about 20:10 and 21:38 UTC, many API requests failed with server or rate-limit (429) errors, and responses were slower than usual. This was caused by a network issue at one of our infrastructure providers. Error rates have been back to normal since 21:38 UTC. We’re monitoring while we work with the provider to prevent a recurrence. --- [2026-11-03T21:12:44Z] (investigating) Since about 20:10 UTC, many requests to the API are failing or returning rate-limit errors, and successful requests may be slower than usual. We are working with our infrastructure provider and moving traffic to restore service. We will post another update within 30 minutes.

**Elevated rate-limit errors and latency on the API** Between 09:32 and 10:02 UTC, and again between 16:47 and 17:08 UTC on October 22, some API requests returned rate-limit (429) errors, and responses were slower than usual. Both were caused by a routing issue in our inference service under heavy load. The API recovered after each window and is operating normally. We’re working on a permanent fix. --- [2026-10-22T17:17:30Z] (monitoring) **Monitoring:** Between 16:47 and 17:08 UTC, some API requests returned rate-limit (429) errors, and responses were slower than usual. A similar issue occurred earlier today, between 09:32 and 10:02 UTC. Both were caused by a routing issue in our inference service under heavy load. The API is performing normally again, and we’re monitoring while we work on a permanent fix. --- [2026-10-21T19:37:49Z] (monitoring) We have applied a mitigation and API error rates and latency have been back to normal since around 19:25 UTC. We are monitoring closely while we work on a permanent fix. --- [2026-10-21T19:23:48Z] (identified) Some API requests are failing or responding slowly. We have identified the cause and are working on a fix. We will post another update within 30 minutes.