Here\u2019s a more complete comparison:<\/p>\n This puts the model more or less at the frontier of agentic coding benchmarks, with only ~750B parameters \u2013 a third of Kimi K3! The Z.ai blog post is rather straightforward, and starts with a bold sentence:<\/p>\n Scaling post-training is all we did for GLM-5.3.<\/p>\n GLM-5.3 is the same base model as GLM-5.2 with substantially extended post-training. To risk a broad oversimplification, Z.ai seems to have a strength in post-training when compared to Kimi, which is more of a pretraining masterpiece. Following this release there have been a lot of discussions wondering how China can keep up so well? How can such a small model be matching the leading public American models? Are these results real?<\/p>\n The simplest explanation is that Z.ai is very good at what they do \u2013 it\u2019s worth recalling that they\u2019ve been working on this line of models longer than almost anyone in the industry. Here\u2019s a brief history of the GLM models.<\/p>\n GLM 5.2, released on June 22 of this year, was a big deal<\/a> \u2013 weeks after the release, I regularly heard from AI researchers I know who still used the model due to its speed (some deploy the model on internal clusters for faster speeds than public offerings) and simplicity (as a model with no rollbacks, etc., when working on frontier AI systems). GLM-5.2 altogether stood up to the hype.<\/p>\n I\u2019ve been going through some of the same denial myself, thinking \u201chow do they keep doing this? Surely the models aren\u2019t as good as they look.\u201d There\u2019s something a bit off-putting with how the American companies have such a commanding resource lead, but can\u2019t seem to pull away in capabilities. The common answer is distillation, which I\u2019ve written<\/a> at length<\/a> about, but I deem not to be the major factor. On that note, there was a recent paper<\/a> that showed simple methods for extracting the reasoning traces from frontier models \u2013 this is the sort of thing that Chinese labs could definitely use at scale. I\u2019m confused why the labs in the U.S. haven\u2019t patched this behavior faster; instead they\u2019re running to the government asking for policy help. It doesn\u2019t add up for me.<\/p>\n Z.ai\u2019s blog is direct and matches with an RL-dominated training regime. They say they used \u201cmore environments, more diverse tasks, and more compute spent training on them.\u201d One does not simply \u201cdistill\u201d RL environments, infrastructure to run them at scale, or algorithms to mix them together effectively.<\/p>\n So, how do the Chinese labs do it if not distillation? Are they benchmaxxing? An accepted definition of benchmaxxing is focusing the model on the test sets, such that the real-world performance meaningfully differs from the on-paper scores. The determining factors are much more big picture than technical (yes, the technical details definitely matter, but are harder to differentiate from lab to lab):<\/p>\n The time to release for Z.ai is likely days, not months as with OpenAI or Anthropic. It is very, very likely that OpenAI and Anthropic have far better internal models than Z.ai and Moonshot AI. Still, these American companies tend to take months to release their models to the public<\/a>, which massively flatters the Chinese labs in adoption decisions at the frontier. To put it simply \u2013 the Chinese labs use all the time that American labs do pre-release testing to keep hillclimbing on benchmarks (SpaceXAI is likely far closer to the Chinese labs here). With the pace of progress being so fast, this is likely the largest determining factor of why Chinese labs stay at the frontier. This, so far, has been economically acceptable for the American labs, as they\u2019ve still had massive demand for their models.<\/p>\n As model self-improvement loops ramp up within the labs building LLMs, if any of these feedback loops require user data, this faster release cycle could massively favor the Chinese labs, giving their offerings longer lifespans before the next vastly superior model comes out, undercutting demand for their models.<\/p>\n These are very clearly the race dynamics that many in the industry worry about. With so many labs building frontier models in the envelope of leading capabilities, it is hard to see this abating in the near future.<\/p>\n Yes, Z.ai probably cares slightly more about public benchmarks than OpenAI or Anthropic. These benchmarks, e.g. scoring highly on the Artificial Analysis Intelligence Index, or similar aggregators, have a very direct impact on their stock price. They in many ways need to do this to keep raising capital and maintain team morale, as being the scrappy underdog matching American giants is a wonderful story. <\/p>\n Subtle benchmaxxing does not need to come out of desperation or any similar pressures. It\u2019s the industry standard across a remarkable number of labs. Many companies\u2019 data acquisition strategy is to buy data on the benchmarks they\u2019re behind on.<\/p>\n Z.ai is not benchmaxxing to the point where GLM-5.3 is fried (at least not intentionally, and they\u2019ll check for it). Every lab is dealing with the rough edges of scaling RL right now. Anthropic\u2019s Opus 5 and Sonnet 5 models have very mixed reputations, despite the incredible benchmark scores. Everyone in the industry is in the same boat, so some model weights end up being easier to use than others, but the benchmark scores in their release blogs are the real deal.<\/p>\n GLM-5.3 is likely a narrower model than Claude Fable or GPT Sol. When GPT-5.2 was released, it had mixed reviews outside of agentic coding. At the same time, OpenAI and Anthropic support very large businesses with countless use-cases for their models. This is a benefit of being a company earlier in their adoption curve \u2013 you can target the most valuable use-cases. Within post-training, caring about a bit less will make assembling the final model far easier.<\/p>\n I\u2019m overstating this a bit, as Z.ai reportedly reached $1B of ARR<\/a> on the back of a strong on-premises deployment business.<\/p>\n Similarly, the flagship GLM models have not had visual capabilities. Being text-only definitely helps Z.ai get more competitive scores, but it is a more competitive space. On the other side of things are models like Inkling-Small<\/a>, which is designed to be omnimodal.<\/p>\n![\"\"](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2026\/08\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/49ade27f-0d84-40f3-840a-3c919e0bb8af_4239.jpeg\")
<\/a><\/p>\n![\"\"](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2026\/08\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/8cf91f6a-1a91-49b9-b220-f3324f2d90b4_1782.jpeg\")
<\/a><\/p>\n