GPT 6.1 Terra reportedly cuts factual error rate from 22.3% to 4.3% on internal hallucination bench
JeremyNguyenPhD · x · 2026-10-27
A claim circulating on X says OpenAI's GPT 6.1 Terra delivers a massive hallucination improvement, cutting its factual error rate from 22.3% (5.6 Terra) to just 4.3% — a reduction of nearly 81% over its predecessor.
Caveat: when Jeremy Nguyen PhD asked for details, OpenAI's site only describes the hallucination benchmark as "internal," so the numbers currently can't be independently verified.
Related event: GPT-6.1 Terra Slashes Factual Error Rate from 22.3% to 4.3%(3 posts)→
More from Models
- UC Berkeley paper: RL reweights LLM reasoning strategies rather than expanding them — burny_tech · 2026-10-27
- Artificial Analysis launches AA-3D Explorer to compare model-built 3D scenes in your browser — ArtificialAnlys · 2026-10-27
- ChatGPT keeps rejecting innocent game-art prompts as 'harassment and bullying' — AnonymousForALittle · 2026-10-27
- Qwen3.5-4B Fine-tune Specialized in Three.js 3D Code Trends on Hugging Face — ccore · 2026-10-27
- Steering vectors turn Qwen3.5-4B into "Golden Gate Qwen" with a known recipe — burny_tech · 2026-10-27
- Experiment finds 9 'gender neurons' in Qwen3.5-4B that flip its self-identity — burny_tech · 2026-10-27