Leaked: OpenAI's Astra solved 10 open math problems, hits Critical cyber threshold
johnseach · x · 2026-09-02
A rumor roundup on OpenAI's next-gen model Astra (unconfirmed):
- Positioning: Astra is the next model family after GPT-5.6 Sol / Terra / Luna.
- Reasoning: an internal build reportedly solved 10 long-open problems in maths and theoretical CS with Lean-verified proofs—pointing to stronger scientific reasoning and long-horizon multi-agent work, agents staying on one hard problem for hours or days rather than a single chat turn.
- Cybersecurity: reportedly the first OpenAI model at the Critical threshold—finding previously unknown vulnerabilities and developing working exploits on hardened systems without step-by-step human guidance; 100% on ExploitBench and two zero-days in internal tests.
- Launch expectations: no exact date, name, or pricing yet. Advanced cyber tools go to testers first, then Daybreak Blue partners (defensive use); the public ChatGPT/Codex/API version will carry extra friction—reviews, pauses, tighter refusals; the system card should give the real eval picture.
- Watch for: whether it ships as GPT-6, a GPT-5.x Astra tier, or its own name; how much of the long-running agent stack reaches normal users; how quickly safety friction relaxes; and whether the maths-style reasoning shows up in coding and science workflows the way Sol did for efficiency.
Related event: OpenAI Teases Astra, First Model to Hit Critical Cybersecurity Threshold(20 posts)→
More from Models
- Qwen3.8-Max-0902 Launch: Enhanced Coding & Agents at $2 Input — Alibaba_Qwen · 2026-09-02
- User praises Fable 5.1, puts pressure on OpenAI's Astra — ns123abc · 2026-09-02
- Dev tempted to switch back from Qwen 3.8 to 3.6: great coder, terrible collaborator — Chuyito · 2026-09-02
- Test Shows Claude Fable 5.1 Finding Bugs Missed by Other Models — Sauers_ · 2026-09-02
- Fable 5.1 'pure insanity': user says it one-shotted his monthly usage limit — talkaboutdesign · 2026-09-02
- OpenAI's Astra rumored to use latent space reasoning, moving beyond text chains — Crazyscientist1024 · 2026-09-02