GLM-5.2: China's "Claude Killer" or a Promising Competitor? My Analysis
A serious intrigue is brewing in the AI market. The Chinese company Z.ai has released the GLM-5.2 model, which many have already rushed to dub the "killer" of Anthropic's flagship Claude. Is there a real threat to market leaders behind this hype? Let's figure it out without unnecessary emotions.
GLM-5.2 is not just another language model. Its main trump card is a giant context window of 1 million tokens, which is five times larger than that of its predecessor GLM-5.1. This allows processing huge volumes of code or documentation in a single session without loss of quality. The model offers two levels of "reasoning enhancement" (High and Max), is distributed under the open-source MIT license, and supports self-hosting. The API price has remained at the level of the previous version, which in itself is a strong argument.
Benchmarks: Where is the truth, and where is the advertising?
According to the developers, GLM-5.2 is the strongest open model on the market. However, comparing it with the proprietary monster Claude Opus 4.8 must be done carefully. On the Terminal-Bench 2.1 test, it scored 81.0, closely approaching Opus 4.8's 85.0 and surpassing Gemini 3.1 Pro (74.0). On SWE-bench Pro, the gap with Opus 4.8 is already 7 points (62.1 vs. 69.2), and on NL2Repo, it is over 20 points (48.9 vs. 69.7).
The most interesting indicators are on long-horizon tasks. On the FrontierSWE test, where the model works for tens of hours, GLM-5.2 lags behind Opus 4.8 by only 1%, while surpassing GPT-5.5 and the previous version Opus 4.7. On SWE-Marathon, the gap with the leader is 13%, but this is still the best result among open models.
Practice vs. Theory: What Users Say
Initial user reviews paint a mixed picture. The model's strengths are acknowledged: excellent basic logic, comparable to GPT-5.5 at a high reasoning level, the ability to autonomously correct errors, and persistence in achieving goals. However, criticism is just as abundant.
The main complaints are about the cloud infrastructure. Users complain about the high cost of plans (from $12.6 to $112 per month), weak support, and the model's tendency to "get stuck" in endless loops, ignoring commands. Many note that GLM-5.2 only truly shines in Max mode, which burns through tokens much faster. There is an opinion that the model is tailored exclusively for benchmarks, not for real-world code.
My analysis: GLM-5.2 is undoubtedly a breakthrough for the open-source AI segment. It demonstrates that Chinese developers can create competitive solutions approaching the quality of top Western models. However, calling it a "Claude killer" is premature. It lags behind Opus 4.8 in most tests, and issues with infrastructure and usability make it more of a tool for enthusiasts and researchers than a ready-made commercial product. The model will only pose a real threat to leaders when Z.ai can offer a stable service at an adequate price with proper support. For now, it is a promising but raw competitor that is closing the gap, but not closing it entirely.