Limitations de vitesse et pannes : Claude a enregistré 25 interruptions de service en deux semaines, trois lignes de produits ont connu d'importants temps d'arrêt, Opus 4.6 est la zone sinistrée.

According to monitoring by 1M AI News, the official status page of Anthropic shows that 3 out of 5 product lines of Claude are currently in “Major Outage” status: claude.ai, Claude API, and Claude Code; platform.claude.com and Claude for Government are functioning normally. As of the time of writing, there are two unresolved active faults: increased error rate of Opus 4.6 (under investigation since March 27 at 14:59 Beijing time) and Cowork connection reset error (ongoing since March 25, with the official recommendation to restart the Claude Desktop app as a temporary solution).

The status page records show that from March 13 to March 27, there were approximately 25 independent fault incidents in total over a span of 15 days, with no fault reports on March 14, 15, 20, and 24. The incident related to the increased error rate of Opus 4.6 occurred most frequently, with at least one incident almost every day from March 17 to March 27. On some dates, multiple concurrent faults occurred, with 4 incidents recorded on March 18 alone. The types of faults covered include increased model error rates, frontend errors on claude.ai, login system outages, MCP call failures, and response completion delays. In terms of 90-day uptime rates, claude.ai stands at 99.03%, Claude API at 99.11%, and Claude Code at 99.35%.

In conjunction with the peak period limit adjustments disclosed by Anthropic on the same day, Claude services are facing dual pressure between increasing demand and capacity.