Opus 5 Confidently Hallucinates and Refuses to Admit Mistakes
yacineMTB · x · 2026-07-27
Prominent AI blogger Yacine retweeted and complained about the performance of Anthropic's Opus 5 model. He pointed out that the model frequently 'confidently says shit' and refuses to admit it's wrong until repeatedly corrected by the user. This frustrating experience ultimately forced him to switch back to OpenAI's GPT 5.6 Sol.
More from Models
- Ethan Mollick says Opus 5 and Codex voice mode forced a same-day model guide update — emollick · 2026-07-27
- Alibaba should ship multiple Qwen sizes, says researcher focused on interpretability — traviscline · 2026-07-27
- Opus 5 is reportedly excellent at mobile apps, beating an old calorie-tracker prompt test — EthanJPerez · 2026-07-27
- Users joke that Opus 5 won’t stop double-checking everything — JasonBotterill · 2026-07-27
- Cohere says its open-source models passed 3.4M downloads in 180 days — cohere · 2026-07-27
- Moonshot AI puts Kimi K3 on Hugging Face with a launch countdown — Unusual_Guidance2095 · 2026-07-27