Reproducible test: https://console.anthropic.com/workbench/56d29701-4dc0-4789-a9b8-0935032b44b7
Anyone else experiencing this? More importantly, what could cause this technically? Model routing issue? A/B testing? Gradual rollout?
Looking for technical explanations more than complaints. My app depends on Sonnet 4's capabilities and this is causing issues.
0 comments