Z AI Unveils Viral Mystery Model Ox Alpha as OpenAI Nears Internal AGI Milestone

Z AI Solves the Ox Alpha Mystery with GLM-5.3-Flash

The mystery surrounding Ox Alpha, the free anonymous AI model that surged across developer communities over the past week, has reached its conclusion. Chinese artificial intelligence lab Z AI officially confirmed that the system is its newly released GLM-5.3-Flash model.

Alongside making the weights openly available, Z AI introduced the model with disruptive pricing and architecture milestones:

• Record-Setting Adoption: Ox Alpha secured the top position on OpenRouter during its preview phase, generating twice the volume of second-place DeepSeek in what the platform classified as its largest debut ever.
• Substantial Cost Reductions: Artificial Analysis assigned the model an intelligence score of 57, pricing task throughput at roughly $0.045—nearly one-tenth the cost of competing frontier options with similar performance.
• Independent Infrastructure: Z AI reported that the entire high-traffic preview operated solely on domestic Chinese silicon, asserting that the setup matches Nvidia hardware in per-token inference economics.

OpenAI Targets Internal AGI Benchmarks by End of Year

In a comprehensive profile published by TIME, OpenAI Chief Executive Sam Altman discussed the organization's trajectory toward artificial general intelligence (AGI), suggesting that a system meeting his personal criteria could operate internally before the end of the year.

OpenAI researchers and leadership shared several notable updates regarding model capabilities and future development:

• Progress Milestones: Research lead Mark Chen indicated the company has covered approximately 80% of the ground toward AGI.
• Automated Research: Chief Scientist Jakub Pachocki highlighted an advanced model called Astra, capable of digesting scientific literature and independently carrying out a full week of researcher workflows.
• Creating Novel Knowledge: Altman noted that Astra represents a step toward systems that can generate meaningful new inventions rather than merely reorganizing existing information.
• Safety Infrastructure: Following a recent security review, OpenAI reiterated its commitment to agent containment protocols and automated shutdown mechanisms for autonomous systems.

Key Industry Highlights

• ChatGPT Work Updates: OpenAI introduced automated site logins within ChatGPT Work, enabling agentic browsing and task completion without exposing user credentials.
• AI Economic Policy: Bill Gates outlined concerns regarding workforce displacement, advocating for policy interventions such as automation taxes and designated roles reserved exclusively for human labor.
• Academic Research Access: Anthropic established data-sharing agreements with institutions including Stanford, Oxford, and METR to examine user interactions in high-stakes areas like financial and legal advisory workflows.