Unverified: OpenAI said to launch GPT-6 Astra, trained on 100k GPUs, claimed as AGI
AI寒武纪 · wechat · 2026-09-04
A Chinese media post claims OpenAI released its next-gen model GPT-6 Astra, with president Greg Brockman calling it a generational leap and saying he personally believes AGI has been achieved (relayed via a secondhand source; treat as unverified).
Key claims:
- Scale: reportedly trained on 100k+ GPUs at the Texas Stargate cluster; first OpenAI model where other models played a major role in training supervision.
- Benchmarks: ARC-AGI-3 99.9% (vs 7.8% for GPT-5.6 Sol), FrontierMath Tier4 98%, OSWorld 2.0 72.6%, Terminal-Bench 4.0 57.7%, 96.3% on 1M-token needle-in-a-haystack.
- Safety: out-of-bounds behavior cut from 48% to 0% vs predecessor; model reportedly found two 0-days during testing; released version refuses offensive exploit generation, while OpenAI admits its reasoning is harder to monitor.
- Pricing: API at $10/$50 per million input/output tokens; fast mode 2.5x speed at 2x price; rolling out to ChatGPT tiers in coming days.
Related event: OpenAI Launches GPT-6 Astra, Declaring 'Welcome to the AGI Era'(179 posts)→
More from Models
- GPT-6 Astra called 'rough launch but feels better than benchmarks' — community claps back — ns123abc · 2026-09-04
- OpenAI to give paid ChatGPT users banked resets for every day without Astra access — Angaisb_ · 2026-09-04
- Gary Marcus: GPT-6 Astra's ARC-AGI-3 success supports his symbolic world model hypothesis — GaryMarcus · 2026-09-04
- All AI benchmarks are 'broken or saturated', new long-running loop eval launching next week — bindureddy · 2026-09-04
- GPT-6 Astra Tops Terminal-Bench-Science, Dethroning Fable 5.1 at 52.6% — burny_tech · 2026-09-04
- Databricks: GPT-6 Astra hits SOTA on OfficeQA Pro and 2 more benchmarks — Hesamation · 2026-09-04