Archived story — belongs to the Friday, September 11, 2026 sweep.
P1 · Tectonic · Frontier AI · sweep Friday, September 11, 2026
Launch of GPT-6 Astra and its impact on cost per task
OpenAI releases GPT-6 Astra with substantial improvements in computer use, coding and reasoning (62.7% on ARC-AGI-3), at 2.5x the price per token but with drastically lower cost per task, which reconfigures the deployment economics of autonomous agents and pressures competitors such as Anthropic and Google. Immediate enterprise adoption (Playco, Legora) and the emergence of alternative harnesses that raise the score to 99.9% show that competitive advantage is shifting from the base model to evaluation and orchestration infrastructure.
Title, thesis and gravity assessment written by an AI model, without prior human review.
Gravity85/100 · Tectonic
Why this gravity
OpenAI's GPT-6 Astra enters as a frontier model with SOTA in computer use and coding, 62.7% on ARC-AGI-3 and drastically lower cost per task despite 2.5x per token, with immediate enterprise adoption by Playco and Legora and direct pressure on Anthropic and Google. It is a structural industry move, but does not validate AGI nor break a benchmark in an incontestably historic way.
- Independent sources
- 5
- Cross verticals
- Frontier AI, Tech Industry, Investing
- Tier 1 actors
- OpenAI, Anthropic, Google
- Gravity vs. the previous edition
- no change (85)
Lead source
OpenAI's GPT-6 Astra Is Shockingly Good at Almost Everything · Decrypt / Jose Antonio LanzMore: Simon Willison, Latent Space, OpenAI Blog, Techmeme
Timeline · Madrid time
- 14:00
- 14:00Legora reviewed 41 documents in minutes with GPT-6 Astra · OpenAI Blog
- 22:05
- 22:18GPT‑6 Astra · Simon Willison
- 23:09
- 07:18
- 01:59The Pelican comparison grid for Astra is pretty interesting · Simon Willison
- 01:27Introducing GPT-6 Astra for developers · Simon Willison
- 19:01
ShareWhatsApp