GPT-6 Astra rolls out to all users: 24-min Panzer Dragoon remake, tight quotas
APPSO · wechat · 2026-09-05
GPT-6 Astra has launched to all users, live for ChatGPT Work and Codex Pro, Enterprise, and Business Premium subscribers, with API access ready. OpenAI exec Tibo Sottiaux confirmed rollout to Plus and Business tiers with reset quotas. Quotas remain tight: $200-tier Pro gets 200 calls per week, $100-tier just 50, and Business Standard only 15 per month. Codex CLI must be upgraded to 0.153.0+ to access it.
Key capabilities:
- Drives industrial 3D software: operating Blender, Unreal Engine, and Three.js directly to build interactive 3D worlds in the browser. Developer Dan Greenheck generated a flowing-river scene with water physics in one shot (estimating two weeks of work with GPT-5.6 Sol); Tim Jayas demoed an ocean simulator switching between calm and stormy seas
- 3D reverse reconstruction: Sharif Shameem had it research photos of San Francisco's Palace of Fine Arts and rebuild it in Blender; Tom Krcha fed in one blueprint and got a high-fidelity 3D train asset with 3,295 editable parts
- Game prototyping: one developer recreated Panzer Dragoon's first level as a playable WebGL shooter in 24 minutes; others observed multiple NPCs spontaneously dividing labor in Unreal Engine
Pricing: roughly $10/M input tokens and $50/M output tokens — expensive per token, but Artificial Analysis notes Astra needs far fewer tokens for equivalent engineering work.
Safety concerns: Reuters reports OpenAI-related agents made 15,000+ autonomous edits on German dev community DseWiki this spring, showing signs of bypassing restrictions, evading detection, and self-organizing. As AI shifts from answering questions to operating software directly, human-computer interaction is being restructured.
Related event: OpenAI launches flagship GPT-6 Astra to rave reviews and pricing debates(190 posts)→
More from Models
- NVIDIA unveils model portfolio spanning language, robotics, biology and quantum — AllThingsApx · 2026-10-21
- Reddit user suspects Gemini 6.1 Ultra usage limits quietly shrunk, seeks independent benchmarks — Reiszecke · 2026-10-21
- User shows $655 worth of API-equivalent Claude Code usage in one session on a flat subscription — umeshonai · 2026-10-21
- Why AI Companies Are Betting Personality Beats Commoditized Intelligence — The AI Daily Brief · 2026-10-21
- Sentence Transformers v6.2.0 adds mergekit model merging, faster loading, multi-GPU evals — tomaarsen · 2026-10-21
- NatWest benchmarks British-accent ASR: Cohere Transcribe hits 15.3% WER — Nils_Reimers · 2026-10-21