# status
RuAPI status: model reliability over 30 days
The share of successful real-user requests for each model family, and how each model did day by day. Major models have backup channels: if the main channel fails, the request automatically moves to a backup — the stats count the final result.
Over the last 30 days, 99.81% of RuAPI user requests succeeded: Claude — 99.79%, GPT — 99.14%, Gemini — 99.4%, DeepSeek — 100%, other models — 100%. 847 requests were completed despite a provider error, thanks to an automatic retry or a switch to a backup channel. Errors caused by the request itself (invalid parameters, unreachable image links, moderation refusals) are not counted. Updated 2026-10-27 14:56 UTC.
Loading data…
How we count
- We use real user requests from the last 30 days, not synthetic checks. Test accounts are excluded.
- A request counts as successful if the user got a response in the end — including after an automatic switch to a backup channel.
- Errors caused by the request itself are not counted: invalid parameters, unreachable image links, oversized requests, moderation refusals. Rejections due to balance or key limits are not counted either.
- Failures include provider errors and timeouts, provider-side rate limits, and cases where no channel was available for the model.