Anthropic warns Opus 5 may leak fake tool calls when thinking is off
heypearlai · x · 2026-07-29
Opus 5 can silently break agent loops without thinking
The thread says that with thinking disabled, Opus 5 may emit a tool call as plain text instead of a real call. In that case nothing executes, the loop takes no action, and the fake call remains in history as if it had happened.
It also warns that the same mode can leak internal tags into visible output. According to Anthropic, telling the model never to output those tags can make the leak worse.
Recommended fix:
- keep thinking enabled
- lower the effort level instead of disabling thinking
The post also notes that Opus 5 can narrate its own corrections mid-response, so structured outputs should explicitly ask for only the final answer.
Related event: Claude Opus 5 Prompting Risks and Best Practices(2 posts)→
More from coding & agent
- Tagvico 3.2.5 checks whether a selected model can really make the app’s tool calls — its_artur1 · 2026-07-29
- Caveman trims AI coding replies to fewer tokens without losing technical detail — KhuyenTran16 · 2026-07-29
- A microservice catalog test found Markdown, GraphRAG, and MCP graphs fail in different ways — JeremyCMorgan · 2026-07-29
- Composio benchmark finds similar success rates but wide speed gaps across coding harnesses — Teknium · 2026-07-29
- LLM APIs still disagree on the parameter name for output token limits — nuno6Varnish · 2026-07-29
- Open-source dashboard compares x402 and MPP across transactions, volume and users — kleffew94 · 2026-07-29