GPT-6 system card: hallucination improving fast, High rating in cyber and bio
inductionheads · x · 2026-10-08
Resharing OpenAI's October system card for GPT-6 Sol and GPT-6 Luna, Andrew Curran argues hallucination is far from intractable. The card details a global ChatGPT rollout replacing GPT-5.6, stronger jailbreak resistance, reduced deception, Astra safety advances, and High capability ratings in cybersecurity and bio/chemical under the Preparedness Framework.
More from Models
- GPT-6 taught to start answering while it keeps thinking, demo shows — isafulf · 2026-10-08
- Pain axis experiment: local Qwen models self-destruct along a mental-turmoil axis; Gemini writes a song — repligate · 2026-10-08
- Dual CMP 170HX (128GB HBM) runs GLM-5.3-Flash at 384K context, ~90 tok/s with EXL3 — Prudent_Appearance71 · 2026-10-08
- GitHub Copilot embraces local AI at Microsoft Windows launch, touting llama.cpp — film_girl · 2026-10-08
- Aleph Alpha Open-Sources Kolibri, a 78B-Parameter German-Reasoning Model Under Apache 2.0 — lmoroney · 2026-10-08
- User mocks ChatGPT error after latest update: 'guess I'll go touch grass' — TheMoonMidas · 2026-10-08