OpenAI Codex hit 'model at capacity' errors as GPT-5.5 demand spiked
OpenAI Codex users reported widespread "elevated errors" Tuesday morning as GPT-5.5 returned "model at capacity" messages and pushed users back to older models. The incident started at 6:32 a.m. ET; OpenAI Codex lead Thibault Sottiaux acknowledged the problem on X, and OpenAI marked all impacted services recovered by 9:48 a.m. ET.
The outage coincided with Anthropic's parallel Claude incident across Sonnet, Opus, and Haiku, with both labs hitting elevated error rates within hours of each other. The double failure highlights how concentrated AI coding workflows have become on two providers, with no graceful degradation path for paying customers.
View full digest for June 17, 2026