Claude Opus 5 flips 30 times on a probability problem before settling on one answer
scaling01 · x · 2026-07-25
A benchmark screenshot shows Claude Opus 5 struggling with a probability question and repeatedly changing its answer before settling on one.
According to the caption, the model:
- identified two defensible answers
- changed its mind 30 times
- finally committed to one answer
- scored 5/5 for distress
The post is basically a snapshot of uncertainty and instability in model reasoning, not a product announcement.
Related event: Claude Opus 5 Shows Erratic Behavior in Probability Tests(3 posts)→
More from Models
- Kimi K3.1 Pro tops MathArena open models, closing frontier gap to about 6 months — teortaxesTex · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26
- supermemory unveils learner-1, claiming continual learning for any model — julianweisser · 2026-10-26
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26