Anthropic says Claude Opus 5 self-verifies, so extra checks can waste tokens
JeremyNguyenPhD · x · 2026-07-25
- Anthropic’s docs say Claude Opus 5 tends to verify its own work on narrow tasks even without being explicitly told to do so.
- Because of that, prompts that add explicit verification steps or ask for a separate subagent to double-check can cause over-verification and waste tokens.
- Anthropic recommends removing legacy verification scaffolding for narrow tasks, because it reduces token usage without hurting quality.
- The docs also warn that Opus 5 may expand the scope of a task on its own, so narrow tasks should be constrained more explicitly.
Related event: Anthropic Warns Claude Opus 5 Prone to Over-Verification(2 posts)→
More from Models
- GPT-6.1 Terra cuts factual error rate from 22.3% to 4.3%, an ~81% drop — haider1 · 2026-10-27
- GPT-6.1 Terra lands: early users praise speed, value and subagent performance — BenBajarin · 2026-10-27
- Leaked GPT-6.1 Terra benchmarks: cheap but 'boring', still far behind Astra and Claude — Angaisb_ · 2026-10-27
- Artificial Analysis Ships Intelligence Index v4.3.2 With GPT-6.1 Terra Comparison — ArtificialAnlys · 2026-10-27
- GPT-6.1 Terra cuts hallucination rate by 50 points while lifting accuracy 2 points — ArtificialAnlys · 2026-10-27
- GPT-6.1 Terra scores 48 on AA Coding Agent Index at one-third the cost of GPT-5.6 Terra — ArtificialAnlys · 2026-10-27