OpenAI officially released GPT-6 Astra, its next-generation artificial intelligence model, Pre-trained on a massive cluster exceeding 100,000 GPUs.
Astra represents OpenAI’s most intelligent, capable, and aligned system to date. Moving past simple text generation, the flagship architecture is engineered from the ground up to operate digital environments, manage end-to-end workflows, and perform complex reasoning across enterprise software, codebases, and research platforms.
Beyond the Chat Box: Desktop & Browser Execution
The defining feature of GPT-6 Astra is its state-of-the-art computer-use capability. Rather than merely outputting text instructions, Astra directly navigates web browsers, updates enterprise CRM records, executes multi-step calendar management, builds web applications, and performs quality assurance testing autonomously.
- Speed & Latency: On the OSWorld 2.0 benchmark a key metric for real-world desktop operations Astra scored 72.6%, completing complex multi-application workflows in roughly 40 minutes. This reflects a 47% reduction in completion time compared to its predecessor, GPT-5.6 Sol.
- Asynchronous Interaction: When encountering ambiguous instructions, Astra proactively makes logical inference calls or asks targeted, asynchronous questions. In developer environments like OpenAI Codex, the model continues executing non-dependent background tasks while awaiting user feedback.
Unprecedented Benchmark Performance
GPT-6 Astra set new industry records across advanced mathematical reasoning, software engineering, and scientific research benchmarks.
Enterprise Workflows & Massive Context
To handle demanding professional workflows, Astra features a 1.05-million-token context window paired with an output limit of 128,000 tokens.
This expanding context…
Source link
Read Full Article by Toluwanimi Adejumo at innovation-village.com
Source link
