DeepSeek's outage credit is sized by the week before a 4h52m partial failure

DeepSeek returns spend for October 8 to 14 ^[1], and the incident began at 00:11 UTC on October 15 ^[5]. The window ends before the outage starts ^[14]. The credit is sized by what an account spent the week before, so a team that was still prototyping gets little back even if the outage hit its launch ^[16].

Once it lands, the credit is spent first. DeepSeek's pricing page says the bonus balance is deducted before a top-up balance when both exist ^[3]. Bonus balance does not answer HTTP requests, and the dev.to write-up notes that failed requests, retries and support tickets are not part of it ^[13]. The notice as reported did not disclose whether the credit expires ^[4].

The shape of the incident matters more for design. The status page listed three affected components: the chat service, the V4 Pro API and the V4.1 Flash API ^[7]. A router that drops from Pro to Flash on error would have stayed inside the same incident ^[15]. DeepSeek has not said what failed. Its notice, as reproduced by IT之家, says it will remove single points of dependency ^[17]. The dev.to write-up interprets that wording as pointing to something shared underneath ^[18]. We'd read it the same way, which means failover has to cross a provider boundary.

That crossing is cheap to code. DeepSeek's docs say the API is compatible with the OpenAI format at the base URL https://api.deepseek.com, and they show it working with the official OpenAI Node SDK ^[10]. The write-up's fallback is a second client with a different base URL and model, triggered only by the errors an outage produces ^[11]. We think that means timeouts, dropped connections and 5xx responses. A 400 or 401 will fail the same way on the second provider, so retrying it there only adds latency.

The client is the small part of the cost. The larger part is a funded account and a key at the second provider, and a prompt set that has been tested against the fallback model before the morning it is needed.

The outage ran 08:11 to 13:04 in Beijing, and fell between 17:11 and 22:04 on October 14 in California ^[12]. On the status page, July to October 2026 uptime was 99.74% for V4 Pro and 99.47% for V4.1 Flash ^[9]. Caixin counts 13 faults since September, four of them partial outages of web and API ^[8]. These provider-wide figures describe your app only if your traffic hits the same models at the same hours.