Meteorite logoMeteorite
Back to videos

Why GLM-5.2 is a bigger deal than you think

On June 12 the US government switched off the most powerful AI model in the world. Fable 5 and Mythos 5, pulled worldwide overnight by an export control. Then on June 17 a lab in Beijing released GLM-5.2, an open weights model that lands within about 1% of Opus 4.8 on coding, and made that kind of switch-off almost pointless. Maybe permanently. The question I keep seeing is why you should care if you're never going to drop tens of thousands on the hardware to run it yourself. That's the whole video. Open weights means a near frontier model can now sit on a hard drive that no company and no government can switch off. There's a floor under the entire thing that didn't exist a week ago, and it's there whether you ever touch the model or not. It also matters if you only ever pay for Claude Code or Codex. A free model this good, at a fifth to a sixth of the price, is the thing keeping the frontier APIs honest. And as these open models keep getting better, which they will, the value stops being about who has access and starts being about what you actually build with it. People are calling this the ChatGPT moment for open source models. I think it's closer to the newsletter moment, and I break down exactly what I mean by that. CHAPTERS 0:00 They switched off the best AI in the world 0:24 Why should you care if you can't even run it? 0:46 Where GLM-5.2 actually ranks 1:55 The one catch: it can't see 2:53 Why it somehow costs more per task 4:28 "It's just a rich person's toy" 5:06 Back to the model that got switched off 5:52 You don't own this stuff, you rent it 6:36 The permanent underclass meme 7:20 Why this matters even if you never touch it 8:41 What's actually keeping your AI bill down 9:29 The value isn't access anymore 10:51 Models commoditise labour, not building 11:11 The newsletter moment for AI 12:02 The door is open now SOURCES & FURTHER READING GLM-5.2 benchmarks and cost per task (the graph in the video): https://deepswe.datacurve.ai/ Z.ai's official GLM-5.2 announcement: https://huggingface.co/blog/zai-org/glm-52-blog GLM-5.2 beating GPT-5.5 at a sixth of the cost: https://venturebeat.com/technology/z-ais-open-weights-glm-5-2-beats-gpt-5-5-on-multiple-long-horizon-coding-benchmarks-for-1-6th-the-cost The Fable 5 and Mythos 5 export control (Anthropic): https://www.anthropic.com/news/fable-mythos-access The 1990s crypto wars precedent: https://www.newamerica.org/cybersecurity-initiative/policy-papers/doomed-to-repeat-history-lessons-from-the-crypto-wars-of-the-1990s/ SpaceX acquiring Cursor for $60 billion: https://www.cnbc.com/2026/06/16/spacex-spcx-cursor-acquisition-ipo.html I'm planning a full video on partial autonomy and why these models commoditise labour, not building. Let me know if you want it in the comments. Follow me on X: https://x.com/robj3d3 Follow me on Insta: https://instagram.com/robhallam_ Subscribe to my free newsletter (weekly tips from founders using AI): https://getmyfirstdollar.com

Suggested videos