DeepSeek outage — not V4 — tops 2026's first hot search; service restored after hours-long interruption

Outage and fix

DeepSeek (深度求索) drew 2026’s first trending search this week — not for a V4 launch, but because its web and app service faltered. Users reported login failures and interrupted conversations on the evening of March 29; the company’s status page says engineers began probing the issue at 00:20 on March 30, issued fixes at 01:24 and 09:13, and declared the problem resolved at 10:33. The timeline is messy: monitoring showed repeated updates and an interruption that lasted nearly nine hours in some measures, while DeepSeek’s own incident record labels the event a major fault with a 7-hour-13-minute duration.

Technically, users were able to log back in after the second update but could not get the assistant to produce answers. Instead, the model reportedly displayed chain-of-thought traces without final outputs. “It may be that the tag marking the end of the thought chain changed format, causing front-end parsing to fail,” one large-model practitioner told Huxiu; it has been reported that DeepSeek’s status posts did not disclose a root-cause and insisted API access was unaffected.

Why it matters

The outage thrust DeepSeek back into the spotlight just as the market buzzes about an imminent V4. DeepSeek rose to prominence during China’s 2025 Spring Festival with its R1 model and rapid V3-series iterations; it has been reported that the company’s V3.2 beta and aggressive API pricing helped trigger a broader price squeeze in the global large-model market. Domestic compute suppliers have reportedly been preparing Day‑0 native adaptations for V4 — but few outside the inner circle can confirm timing. Will a stability hiccup accelerate or delay a big reveal? The question hangs over a sector already under strain from intense global competition and shifting trade and export policies.