User comparison: Gemini nailed an insurance-law question that ChatGPT defended with circular reasoning
Hatrct · reddit · 2026-09-18
A user's side-by-side test on why courts permit sex-based auto insurance rating but restrict medical-condition exclusions: Gemini quickly cited the relevant laws and conceded the user's critique, while ChatGPT got the facts wrong and resorted to circular reasoning until repeatedly pushed. The author concludes Gemini currently performs better on this kind of nuanced reasoning — a single anecdote, but a notable behavioral contrast.
More from Models
- Report: Hackers Used a Loosened-Guardrail Opus 5 to Breach OpenAI's Internal Monorepo — teortaxesTex · 2026-09-18
- NetEase Youdao Open-Sources Confucius4-R2T2, a Streaming ASR That Never Rewrites Committed Text — rohanpaul_ai · 2026-09-18
- Fine-tuned 4B model as a decision scorer with temperature-scaled confidence — Gradio · 2026-09-18
- Users reverse-engineer Astra's ASCII art: likely programmatic coordinate painting, not SVG conversion — MoonL88537 · 2026-09-18
- Researcher Apologizes for Misleading Leaderboard Submission, Promises Rule-Compliant Rerun — AlbertQJiang · 2026-09-18
- Noam Brown's Dwarkesh podcast: a capabilities researcher 'freaking everyone out' on safety — Apprehensive_Sand951 · 2026-09-18