OpenAI hit by another outage as ChatGPT, Codex, and APIs go down together
Saturday’s disruption is the fourth service incident in four days for a company whose unified platform now serves 900 million weekly users and 10 million agentic users (July 25, 2026 – 10:54 am)
TL;DR
OpenAI’s APIs, ChatGPT, and Codex all reported elevated error rates on Saturday, marking the fourth service disruption in four days for the company. Users encountered 503 errors with the internal label “biscuit_baker_service_me_circuit_open,” which prevented requests from reaching OpenAI’s servers.
Recent Service Disruptions
- Saturday: All three pillars of OpenAI’s product line simultaneously affected, including 12 affected components in the API, 15 in ChatGPT, and four in Codex.
- Thursday: Elevated API error rates reported.
- Wednesday: Elevated errors observed in Codex Review and both ChatGPT and API image generation.
- Previous months: A pattern of approximately 166 incidents across nine months, averaging roughly 18 per month, starting from autumn 2025.
- April 20: Major outage for ChatGPT.
- July 23: Another disruption just two days before Saturday’s incident.
Reliability Concerns
The frequency of these outages raises reliability questions, especially with OpenAI’s agentic products reaching 10 million weekly users and ChatGPT Work launching this month. The dependence on AI agents for multi-step tasks requires a consistent platform to stay up.
Context
OpenAI is not alone in facing these challenges. Similar concerns were raised when Anthropic’s Claude went down in June, questioning whether the infrastructure behind frontier AI models can meet the demands placed on it. The scale and ambition of OpenAI’s unified platform, serving 900 million users, only amplify the importance of resolving these issues for enterprise buyers considering handing real work to AI agents.
Story by Ana Maria Constantin