OpenAI Unveils GPT-6 Astra For Autonomous Professional Workflows, Pledges $1B To Frontline Cybersecurity Defense
AI analysis firm OpenAI unveiled GPT-6 Astra, its newest frontier mannequin engineered for autonomous laptop use, software program engineering, scientific analysis, and complicated skilled workflows. The system turned obtainable to a restricted set of organizations on launch day and can roll out to ChatGPT Plus, Pro, Business, and Enterprise subscribers, in addition to by means of the OpenAI API and main cloud suppliers, within the coming days.
The firm positions Astra as a frontrunner throughout a number of benchmark classes. It reportedly scores 99.9% on ARC-AGI-3, saturates FrontierMath Tier 4 at 98%, and achieves 100% on ExploitBench. In sensible evaluations, the mannequin reached 59.3% on Agents’ Last Exam, 72.6% on OSWorld 2.0, and 96% on the graduate-level GPQA Diamond science benchmark. OpenAI highlights vital effectivity positive aspects, noting that Astra completes computer-use duties in roughly 47% much less time than its predecessor whereas utilizing considerably fewer output tokens. In terminal-based coding and scientific workflows, it scored 57.9% on Terminal-Bench 4.0 and 64.6% on Terminal-Bench Science 0.1, with estimated API prices under these of competing configurations. Independent researchers at Epoch AI additionally famous a brand new estimated compute effectivity document of 169.
Astra introduces a redesigned context-management mechanism in OpenAI’s Codex surroundings that preserves searchable working notes throughout prolonged classes, avoiding the knowledge loss sometimes related to context-window compression. The mannequin can be designed to deal with multistep skilled duties, together with information evaluation, doc drafting, CRM updates, and specialised engineering work.
Cybersecurity, Alignment, and a $1B Defender Commitment
OpenAI is emphasizing Astra’s security properties alongside its capabilities, describing the mannequin as its most aligned launch thus far. Internal evaluations point out substantial enhancements in user-intent understanding and task-boundary respect. In adversarial assessments, Astra produced misaligned outcomes at a fee of two.4%, considerably under reported figures for competing frontier fashions. The system additionally demonstrated improved resistance to circumventing automated overview mechanisms and a diminished tendency to hallucinate about its personal capabilities.
In cybersecurity evaluations, Astra meets OpenAI’s inner Critical preparedness threshold. It achieved an ideal rating on ExploitBench and a 42.4% success fee on ExploitGym, whereas reportedly discovering two beforehand unknown zero-day vulnerabilities throughout evaluation. Standard deployments will refuse requests to create proof-of-concept exploits, although OpenAI plans to develop defensive entry by means of its Daybreak program within the weeks forward.
Concurrent with the launch, OpenAI introduced a $1 billion dedication to subsidize frontier AI cybersecurity sources for frontline defenders. The Daybreak for Frontline Defenders initiative targets important service operators—together with water utilities, electrical grid operators, native governments, and group banks—beginning within the United States with deliberate worldwide enlargement.
The program includes a pilot with the Multi-State Information Sharing and Analysis Center and integrates greater than 35 companion merchandise by means of the Daybreak Defense Network. API entry is priced at $10 per million enter tokens and $50 per million output tokens, with an non-compulsory quick mode obtainable at double the usual fee. Enterprise directors should explicitly allow Astra for his or her workspaces, and the mannequin helps Zero Data Retention for eligible API clients.
The put up OpenAI Unveils GPT-6 Astra For Autonomous Professional Workflows, Pledges $1B To Frontline Cybersecurity Defense appeared first on Metaverse Post.
