OpenAI released GPT-6 Astra

OpenAI released GPT-6 Astra. the release off the new model coincides with the new release off Anthropic's fable 5.1 ai model and there is a lot off content circling online off some off the great things people have done already with these models , Instagram page metav3rse from the pubity group posted about the differences between the two models and some off the things people have done with the models including building video games .

Key Highlights

Performance & Capabilities: OpenAI hailed Astra as its most advanced model, describing it as entering the "AGI era". It sets new benchmarks in multi-step reasoning, autonomous computer and browser execution, complex software engineering, and mathematical problem-solving.

New Reasoning Architecture: Astra uses a new reasoning technique known as recurrent depth (or "looped transformers"). This technique significantly improves computational efficiency and task handling, though safety researchers have noted it makes tracking the model's internal "chain of thought" more opaque.

Critical Cybersecurity Threshold: Astra scored 100% on the ExploitBench cybersecurity benchmark, crossing OpenAI's critical capability threshold for offensive/defensive cyber operations. As a result, OpenAI restricted full access to advanced security tools, releasing a guarded version to paid tier users.

Rollout Delays & Backlash: Due to high demand and elevated security protocols, OpenAI initially limited initial access to trusted partners and restricted access for ChatGPT Plus and Pro subscribers. The rocky rollout led CEO Sam Altman to issue a public apology over user access restrictions.

With open AI and the ChatGPT AI model brand and Anthropic with the Claude model brand who are both in a national ai race and set for IPO this year , neck to neck and competition is tense .