Early access reviews claim GPT-6 'Astra' is a massive leap at browser control and knowledge work
CurieuxExplorer · x · 2026-09-05
- Matthew Berman says he has early access to GPT-6 "Astra" and calls it the best model he has ever used — a massive improvement rather than an incremental one, especially on zero-shot prompts.
- Its strength centers on knowledge work: slide creation, analysis, writing, and notably browser control, which he says is another level beyond the already strong GPT-5.6 and significantly faster.
- His thread includes playable demos covering games, code, writing, slides, and browser control.
- Note: this is a third-party claim of early access; OpenAI has not officially announced anything, so treat as unverified.
Related event: Early Testers Praise GPT-6 Astra's Browser and Software Skills(2 posts)→
More from Models
- Musk plans to rewrite all human knowledge with Grok and retrain; critics fear data poisoning — JasonBotterill · 2026-09-05
- GPT-6 Astra broken in Cursor BYOK: dev builds relay to bridge Chat Completions and Responses API — salmoukas · 2026-09-05
- Twitter user accuses OpenAI of hiding 'loss of control' incidents from METR evaluators — AaronBergman18 · 2026-09-05
- Claude Code user burns through 50% of weekly usage in six hours, begs for higher limits — sethlazar · 2026-09-05
- Why Models Survive Hostile Training Envs With Persona Changes Firewalled — repligate · 2026-09-05
- GPT-6 Astra reportedly aces Bach chorale benchmark, first model to write passing tones — yacineMTB · 2026-09-05