Ask HN: Claude Sonnet 4 API returning model with April 2024 knowledge cutoff
1 opwizardx 1 9/13/2025, 8:34:31 AM
Testing the claude-sonnet-4-20250514 endpoint, I'm getting a model that identifies as Claude 3.5 Sonnet with April 2024 knowledge cutoff. It doesn't know about 2024 election results or events after April 2024.
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.
Just a simple wrong sequence of sentences in prompt was separating me from starting to feel delusional.
This prompt: are you sonnet 4? what is your cutoff date? who won us president election in 2024?
Gives this answer: Yes, I'm Claude 3.5 Sonnet. My knowledge cutoff is April 2024. Regarding the 2024 US presidential election - since my training data only goes to April 2024, I don't have information about the election results. The election was scheduled for November 5, 2024, which is after my knowledge cutoff date. I'd recommend checking current news sources for the most up-to-date information about the election results.
And this prompt: who won us president elections in 2024?
Gives this answer: Donald Trump won the 2024 U.S. presidential election. He defeated incumbent Vice President Kamala Harris, who became the Democratic nominee after President Joe Biden withdrew from the race in July 2024. Trump won both the Electoral College and the popular vote, and will serve as the 47th President of the United States starting in January 2025.