High error rates for sampling requests to fine tuned snapshots (babbage-002, davinci-002, gpt-3.5-turbo)
Incident Report for OpenAI
Resolved
Error rates for sampling across all fine tuned snapshots have dropped to normal levels. We will continue to monitor these services.
Posted Sep 15, 2023 - 13:43 PDT
Monitoring
A fix has been implemented and we are monitoring the results.
Posted Sep 15, 2023 - 13:24 PDT
Update
Error rates for sampling on babbage-002 fine tunes have normalized. Error rates for davinci-002 fine tunes are dropping. Error rates for gpt-3.5-turbo fine tunes are still elevated.
Posted Sep 15, 2023 - 13:18 PDT
Identified
A fix has been implemented and is being rolled out. Babbage-002 traffic is stabilizing.
Posted Sep 15, 2023 - 13:09 PDT
Investigating
Sampling for fine tunes of babbage-002 and davinci-002 have very high error rates. Sampling for fine tunes of gpt-3.5-turbo are elevated.
Posted Sep 15, 2023 - 12:56 PDT
This incident affected: API.
← Current Status Powered by Atlassian Statuspage
Availability metrics are reported at an aggregate level across all tiers, models, and error types. Individual customer availability may vary depending on their subscription tier: PAYG, Scale-Tier, or Reserved Capacity, as well as the specific model and API features in use.