Full Breakdown
DeepSeek’s October 15 Outage and Full-Week API Refunds
By Drooid · · How we work
Core Event: Massive Service Disruption and Refund Initiative
On October 15, Chinese AI firm DeepSeek experienced an outage that disabled its web chat, mobile app, and API services for the V4.1 Flash and V4 Pro models. The service-status page recorded the incident from 8:11 a.m. to 1:04 p.m. Beijing time (4 h 52 m). DeepSeek later posted an apology and announced that every API customer would receive a credit equal to the amount spent during the prior week (October 8 – 14). Credits were automatically deposited and will apply to future billing without expiration.
Background & Context
DeepSeek’s low-cost, open-source models are widely used, and a Weibo post noted this outage was the latest of thirteen incidents since September, most being degradations. An outage tracker also logged a separate disruption on October 12, suggesting recurring reliability challenges during high demand.
Data & Statistics
- Outage duration: 4 h 52 m per the status page; a later company update described it as “about 1 hour.”
- Fault reports: StatusGator logged 86 incident reports in the preceding 24 hours.
- User-reported issues (U.S. focus): app problems 57 %, website problems 41 %, login issues 2 %.
- Refund amounts: screenshots showed credits from a few yuan up to 3,000 yuan (? $449); several users received refunds exceeding 200 yuan (? $30).
Official Statements & Responses
No detailed technical cause was disclosed.
On-the-Ground Reports
Developers reported halted workflows—article editing, code generation, thesis drafting—as API calls returned errors or timed out. Users in Luxembourg, Poland, and New York saw “server temporarily unavailable” messages and slower responses. Some praised the usage-based refund as more sincere than a uniform voucher.
Why It Matters / Impact
The incident highlights compute strain in China’s generative-AI market, where over 700 million people are active users. By tying compensation directly to API billing, DeepSeek set a precedent for monetary restitution in the Chinese AI sector, prompting debate on whether other providers will follow. The outage also underscored the reliance of downstream applications on a single AI service, raising concerns about systemic risk for businesses that embed DeepSeek models.
Conflicting Reports & Gaps
- Duration discrepancy: 4 h 52 m on the status page vs. “about 1 hour” in a later update.
- Cause details: No specific technical root cause has been identified.
- Geographic scope: International users reported issues, but DeepSeek later claimed systems were operating normally, with no official confirmation of a U.S.-wide outage.
What’s Next
DeepSeek will continue its investigation and implement changes to eliminate single-point failures. Future updates are expected to detail the outcomes of this stability overhaul and any further mitigation measures. Users should monitor DeepSeek’s official status page for notices.
