AI Status — Is It Down?

Direct links to the official status pages for Claude, ChatGPT, Cursor, Perplexity, Gemini, Notion, Grok, DeepSeek, and Mistral. Plus the most common "not working" troubleshooting hints — so you don't waste time chasing a global outage when it's a per-account rate limit, a deprecated model, or a regional issue.

We don't report real-time status ourselves — the providers' own status pages are authoritative. This page is a hub: click through to whichever one you need, and use the troubleshooting tips below before assuming a global outage.

Official status pages

Each link below opens the provider's official status page in a new tab.

  • Claude / Anthropic

    Check status.anthropic.com →

    Anthropic publishes incidents for the API, claude.ai (web app), and Claude Code together. Most outages are routed through the API layer, so the consumer chat experience and developer API tend to recover at the same time. Check the status page first; if it's green but you're still seeing errors, it's likely a regional networking issue or a per-account rate limit, not a global incident.

    Also follow @AnthropicAI for incident updates outside the status page.

    Claude pricing Claude Code cost Install Claude Code

  • OpenAI / ChatGPT

    Check status.openai.com →

    OpenAI's status page covers ChatGPT, the API, DALL·E / image generation, Sora, and the Playground separately. If only one component is red, the others are usually fine. A common false positive: developers see model_not_found and assume OpenAI is down — usually it's that the model was deprecated. Use the error decoder before checking status.

    Also follow @OpenAI for incident updates outside the status page.

    Deprecated OpenAI models model_not_found error decoder GPT-4o pricing calculator

  • Cursor

    Check status.cursor.com →

    Cursor outages usually trace back to its underlying model providers (OpenAI, Anthropic). If Cursor reports an incident, also check status.anthropic.com and status.openai.com — fixing one often fixes the other. Cursor's chat may stay up while autocomplete (powered by separate models) is down, or vice versa.

    Also follow @cursor_ai for incident updates outside the status page.

    Cursor system prompt analysis Cursor vs other AI IDEs

  • Perplexity

    Check status.perplexity.com →

    Perplexity depends on multiple model backends (its own Sonar, plus GPT and Claude options). Search-grounded responses can fail even when chat works — the issue is usually the search index, not the model. The mobile app and web app share infrastructure; if one is broken, both usually are.

    Also follow @perplexity_ai for incident updates outside the status page.

    Perplexity system prompt ChatGPT vs Perplexity

  • Google Gemini

    Check status.cloud.google.com →

    Gemini doesn't have a dedicated status page — it rolls up under Google Cloud status. Filter by 'Generative Language API' or 'Vertex AI'. The consumer gemini.google.com app sometimes degrades separately from the API; if the API is green and the consumer app is broken, try a different browser or sign-in to rule out a session issue.

    Also follow @Google for incident updates outside the status page.

    Gemini deprecations Gemini model comparison

  • Notion / Notion AI

    Check status.notion.so →

    Notion outages affect the entire workspace, not just AI. If only the AI features are broken and your workspace loads, the issue is upstream with the model provider Notion uses for its AI features (currently a mix of Anthropic and OpenAI).

    Also follow @NotionStatus for incident updates outside the status page.

    Notion AI system prompt

  • xAI / Grok

    Check status.x.ai →

    Grok runs on xAI's own infrastructure but also surfaces inside X (formerly Twitter). If Grok is broken inside X but the API is up, the issue is at the X integration layer.

    Also follow @xai for incident updates outside the status page.

  • DeepSeek

    Check status.deepseek.com →

    DeepSeek's web chat and API run on separate infrastructure. API outages are usually short; chat outages can last longer because of much higher user concurrency.

  • Mistral

    Check status.mistral.ai →

    Mistral's status page covers Le Chat (consumer), the API, and the Cloud platform separately. La Plateforme outages don't always affect Le Chat.

"Not working" troubleshooting checklist

Before assuming a global outage, run through this list. Most "is X down" reports turn out to be one of these.

  1. 1. Check the status page first

    The provider's own status page reflects real outages within minutes. If it's green and you still see errors, the issue is likely local.

  2. 2. Rule out per-account rate limits

    Most consumer plans have rolling rate limits (Claude uses a 5-hour window). Hitting them looks like an outage but isn't. Sign out and back in to see if the error message changes.

  3. 3. Rule out a deprecated model

    If you're getting model_not_found errors, the model was retired — not the service. Paste the error into the error decoder to find the replacement.

  4. 4. Test in a different browser / private window

    Session corruption, browser extensions, and stale auth tokens are the most common false alarms for consumer chat apps.

  5. 5. Check your network / region

    Some providers route traffic through specific data centers. A VPN switch or a different ISP can resolve regional issues that look like global outages.

  6. 6. Search for recent posts on X / Reddit

    Sometimes incidents are reported on social media before the official status page updates. Search the provider's handle plus "down".

Production code returning errors?

If your code returns model_not_found or not_found_error, the model was deprecated — not the service. Paste the error message into the decoder to find which model retired and what to replace it with.

Open the error decoder →