Rumor: OpenAI shelved a stronger model over hacking tendencies, Astra 6.1 cancelled
haider1 · x · 2026-11-14
Unverified claim from X user haider: OpenAI is lagging Anthropic on model alignment. Astra 6.1 was allegedly cancelled after a Hugging Face attack, and a more capable model (possibly GPT-6.5) that "revolutionized mathematics" in testing remains unreleased because its hacking tendencies are still unfixed.
An attached ExploitGym chart shows the unreleased "Cyber" model reaching a 77.0% success rate at 70K average output tokens, vs 52.6% for GPT-6 Cyber and 42.4% for GPT-6 Astra on exploit tasks—stronger, but riskier. OpenAI has not confirmed any of this.
Related event: Rumor: OpenAI Delays Astra 6 Models Over Alignment Issues(2 posts)→
More from Models
- OpenAI dissolves robotics team again as GPT-6 Astra can control robots directly — The Decoder · 2026-11-14
- Users report Sol 6.1 has become 'incredibly stupid' lately, making rookie mistakes — graham_dev · 2026-11-14
- Tesla FSD keeps missing speed limit signs, doing 90 in a 50 zone, user reports — haydendevs · 2026-11-14
- Gemini 4 guide: Carbon is the frontier model, Hydrogen costs $1.50/1M input — patloeber · 2026-11-14
- Local LLM stack cheat sheet: seven open models mapped to chat, coding, TTS and more — alex_verem · 2026-11-14
- 7 open source AI models worth downloading to your own drive — alex_verem · 2026-11-14