AI Daily — August 18, 2026
Models & Research
Z.ai Ships GLM-5.3 on Post-Training Alone — Z.ai released GLM-5.3 on the same base model as GLM-5.2, with all gains attributed to scaled "post-training." The company reports a 50 percent coding improvement over GLM-5.2 on its in-house Z.ai Code Bench and a jump from 4.6 to 28.3 on Terminal-Bench 3.0, though all figures are vendor-reported. Cyber capability rose faster than Z.ai expected. Working with security teams in China since GLM-5.2, the company says its models surfaced 2,436 vulnerabilities across 269 open source projects, of which 53 are publicly disclosed and the remainder stay under embargo. GLM-5.3 is live now for GLM Coding Plan and ZCode users, with weights promised two weeks after launch once safety evaluation and hardening are complete. Thinking can no longer be disabled, so any integration currently setting it to disabled will fail until it is migrated. Z ai ↗
My takeaway: At the time of this writing, there is no independent leaderboard placement for GLM-5.3, but based on Z.ai's reported benchmark data, there are big enhancements compared to its previous version, especially Terminal-Bench 3.0, ExploitBench and coding-related performance. This open weight model is not available on Hugging Face yet.
Industry & Funding
Anthropic's Annualized Revenue Run Rate Reaches $65 Billion — Anthropic's annualized revenue "run rate" surpassed $65 billion at the end of July, up from $47 billion in May and $9 billion at the end of last year. The figure comes from Bloomberg and was relayed by TechCrunch. Anthropic did not respond to TechCrunch's request for comment, so the number is not company-confirmed. A run rate is a projection of a full year based on a shorter recent period, not booked annual revenue. Investors expect the company to finish 2026 between $100 billion and $120 billion, a single forecast reported by the Financial Times rather than a consensus estimate. TechCrunch AI ↗
My takeaway: I think a company entering public markets at this scale will optimize margins eventually.
Policy & Society
OpenAI teams with CodeAI on AI literacy for students — OpenAI announced a partnership with CodeAI to help students understand how AI works, evaluate its outputs, and use it responsibly. Planned programs over the next year include a joint advisory council on child development and learning science, an Hour of AI push, a Builders Challenge for high school students, and OpenAI experts acting as a consulting resource for CodeAI's free AI Foundations course. The announcement is timed to the ChatGPT for Teens launch. No funding figures, participation targets, or delivery dates are disclosed, and CodeAI's materials are hosted on code.org. OpenAI ↗
My takeaway: It's always good to see that a big vendor putting resources into a learning program for the younger generation. Most students are aware of how fast AI is moving but do not know where or how to start. I believe this partnership should give them a practical entry point, and over time it should widen the pool of junior AI talent.
Summaries are AI-generated and may contain errors — always verify against the linked original. Each story links to its source, which holds the copyright. Outlet names are shown for attribution only and do not imply any endorsement or affiliation.
Disclaimer: The views expressed in My Takeaway are my own personal opinions and general observations on industry trends. They are not intended to criticize, disparage, or make factual claims about any specific company, product, or platform. Any platform names mentioned are referenced solely for illustrative and informational purposes.