GPT-6 Astra scores 97.6% on FrontierMath but stays gated from regular users for days
burny_tech · x · 2026-09-04
OpenAI has begun its staggered rollout of GPT-6 Astra, which president Greg Brockman calls a "generational leap" bringing the frontier into the "AGI era."
- Availability: starts Thursday with approved cybersecurity defenders in the Daybreak program; paying ChatGPT subscribers and API access follow "over the next several days."
- Official benchmarks: 97.6% on FrontierMath Tier 4 v2, 74.1% on DeepSWE v1.1, 95.9% on BenchCAD, 96% on GPQA Diamond, 100% on ExploitBench, and 98.6% on ARC-AGI-3.
- Outperforms GPT-5.6 and flagship Sol across coding, safety, and computer-use tasks; better at powering autonomous agents and generating docs/spreadsheets from local information.
- VP of research Aidan Clark: largest-ever training run, pre-trained on 100,000+ GPUs.
Related event: OpenAI Launches GPT-6 Astra, Declaring the AGI Era(34 posts)→
More from Models
- OpenAI launches controversial Astra model, starting with Daybreak cybersecurity customers — recoveringasshole0 · 2026-09-04
- GPT-6 Astra lands: 98.6% on ARC-AGI 3, Brockman declares the AGI era — Expert_Annual_19 · 2026-09-04
- OpenAI rolls out GPT-6 Astra to vetted cyber customers at 2.5x GPT-5.6 pricing — rohanpaul_ai · 2026-09-04
- Analyst doubles down: OpenAI set to out-accelerate Anthropic by year-end — scaling01 · 2026-09-04
- GPT-6 Astra reportedly scores 61.2 on AAII, slightly above GPT-5.6 Sol — Angaisb_ · 2026-09-04
- Benchmarks show Claude beating GPT-6 Astra on agentic coding and AA index — burny_tech · 2026-09-04