Anthropic's Claude Code, a cutting-edge AI model, is currently facing a peculiar issue where it returns blank or truncated thinking blocks, failing to provide users with the expected summaries of its thought process. This problem arises even when users explicitly request summarized thinking, raising concerns about the model's functionality and the value users receive for their thinking tokens.
Key Insights
Claude Code is returning empty or truncated thinking blocks for some models, which is currently under investigation. The issue may be an artifact of ongoing tests that are changing how Anthropic displays summaries. Notably, users are still being charged for thinking tokens, even in instances where the thinking text is not returned, highlighting a need for resolution to ensure fairness and transparency in the service provided by Anthropic.
