After the release of GLM-5.2: Vercel CEO was almost shocked. Is open source chasing behind closed source?

If you want to support my video creation, you can consider joining the "SmallWoods" YouTube membership: https://www.youtube.com/channel/UCAVDRj14A9W2Zix1Y5EUm7Q/join - Join my knowledge planet: https://t.zsxq.com/W5Oj7 - Follow me on Twitter: https://twitter.com/verysmallwoods - Follow my Bilibili: https://space.bilibili.com/615957867/ - If you have any questions or suggestions, please contact us by email: 0001coder@gmail.com It has only been two days since GLM-5.2 was released, and the open source circle has been very hotly discussed - and the signal this time is not the benchmark itself, but the CEOs of first-line companies taking the initiative to speak out. This issue talks about: - Programming benchmark (official): FrontierSWE 74.4, surpassing GPT-5.5's 72.6, and less than 1 point behind Claude Opus 4.8's 75.1; SWE-bench Pro 62.1 also surpassed GPT-5.5 - The generation gap with the previous generation: FrontierSWE jumped from early 30s to 74, DeepSWE jumped from 18 to 46 - The most powerful open source weight model now - Real endorsement: Vercel CEO Guillermo Rauch publicly said that he was "almost shocked by its programming ability, This changes things", Box CEO also followed suit - Price and general ledger: It costs a lot of tokens, but MIT is open source and can be deployed locally, and the overall cost is much lower than the closed source flagship - Why is it important: Open source weight + MIT license + first-tier programming capabilities, the first time it appears on a model at the same time - Official blog: https://huggingface.co/blog/zai-org/glm-52-blog - Rauch tweets: https://x.com/rauchg/status/2068517095818809770