GLM-5.2: A Real Threat to Claude or Just Another Chinese Hype?
China's AI sector is once again making bold and audacious claims. The new GLM-5.2 model from Z.ai, according to many enthusiasts, is capable of challenging Anthropic's flagship products. But how justified are these claims? Let's break down the hard numbers and real user experience.
The developers position GLM-5.2 as a flagship model optimized for long working sessions. The key improvement over version 5.1 is a stable context window of 1 million tokens, which is five times larger than the previous figure. This allows the model to keep entire codebases in view without losing quality on ultra-long tasks.
The model offers two levels of reasoning: High for a balance of performance and token consumption, and Max for achieving maximum results, but with increased resource usage. Importantly, GLM-5.2 is distributed under the open-source MIT license, allowing it to be run on your own hardware without regional restrictions.
Numbers and Benchmarks: Breakthrough or Marketing?
Z.ai's own tests are indeed impressive. On key benchmarks, GLM-5.2 shows a significant leap compared to its predecessor. For example, on Terminal-Bench 2.1, the result increased from 63.5 to 81.0, closely approaching Claude Opus 4.8's score of 85.0 and surpassing Gemini 3.1 Pro's 74.0.
On SWE-bench Pro, the model scored 62.1 points versus 58.4 for GLM-5.1, though Opus 4.8 has 69.2 here. In long-term scenarios like FrontierSWE, the gap with Anthropic's leader is only 1%, an outstanding result for an open-source model. However, on tests like NL2Repo and DeepSWE, the gap with Opus 4.8 is more significant—20% and 12% respectively.
Practice vs. Theory: What Users Say
Despite the impressive benchmark numbers, real user experience paints a more complex picture. Many developers note that GLM-5.2 is indeed the strongest open-source model available at the moment. Its basic logic has noticeably improved, and in programming, it is comparable to GPT-5.5 at a high reasoning level.
However, criticism mainly concerns infrastructure and stability. Users complain about weak cloud support, high pricing tiers, and the model's tendency to get "stuck" in infinite loops, ignoring commands. Many note that the model only truly shines in Max mode, which consumes significantly more tokens than High. Ultimately, according to part of the community, it's simpler and cheaper to pay for Claude or GPT.
Cryptalist Expert Opinion: GLM-5.2 is undoubtedly an important step forward for open-source AI. It narrows the gap with proprietary giants, especially in the niche of programming and autonomous agents. However, calling it a "Claude killer" is still premature. Issues with infrastructure, stability, and high token consumption are the "growing pains" that Z.ai needs to solve for the model to become a real alternative, rather than just a bright flash on the industry's radar. For now, it's more of a "Chinese challenge" than a "killer."