Claude Code users are reporting that the model’s thinking blocks sometimes come back blank or truncated, while the reasoning tokens behind them still show up on the bill. The complaints, logged in GitHub issues, question whether developers are paying for reasoning they never see.
The Register reported the issue on Friday. Anthropic attributed the behavior to network tuning, specifically logic that terminates or retries long-running requests. The company says work is ongoing to balance perceived latency against the risk of requests getting stuck.
A spokesperson said Claude Code ships updates nearly every day, and that community reports like these GitHub issues are a big part of how the company catches problems quickly.
The episode lands as Claude Code becomes the centerpiece of Anthropic’s developer push, with rivals like OpenAI’s Codex and DeepSeek’s Harness competing for the same users. Pricing transparency for hidden reasoning is a recurring sore point across agentic coding tools, where chain-of-thought tokens are often billed even when users cannot inspect them.
