GLM-5.2 vs. Claude: Is the New Chinese Model Worthy of the Title "Killer"?
A sensation is brewing in the world of artificial intelligence: the Chinese company Z.ai has introduced the GLM-5.2 model, which has already been dubbed the "killer" of Anthropic's flagship product, Claude, online. My analysis shows that behind this loud title lies a complex and ambiguous picture.
What is GLM-5.2 and what is its strength?
GLM-5.2 is an open-source model with an MIT license, designed for long and complex work sessions. Its main advantage over its predecessor, GLM-5.1, is a context window of 1 million tokens (compared to 200 thousand). This allows loading huge amounts of code or text into a single reasoning cycle without loss of quality. The model supports two levels of reasoning enhancement: High for a balance between performance and token consumption, and Max for maximum analysis depth. Key features:
- 1 million token context without degradation on ultra-long sessions.
- Two levels of reasoning enhancement: High and Max.
- Open MIT license, allowing self-hosting on your own hardware.
- API price remains at the level of the previous version.
Benchmarks: where is the truth and where is the marketing?
According to Z.ai's internal tests, GLM-5.2 is recognized as the strongest open-source model on the market, but it falls short of Anthropic's flagship, Claude Opus 4.8, in most scenarios. The gap is especially noticeable in tests requiring deep code understanding, such as NL2Repo (48.9 vs. 69.7) and DeepSWE (46.2 vs. 58.0). Nevertheless, on the Terminal-Bench 2.1 test, the model scored 81.0, closely approaching Opus 4.8 (85.0) and surpassing Gemini 3.1 Pro (74.0).
On long-horizon tasks, the picture is similar. On the FrontierSWE test, where the model manages open technical projects for hours, GLM-5.2 lags behind Opus 4.8 by only 1%. This is an impressive result, especially considering it outperforms GPT-5.5 and the previous version of Opus 4.7.
Price and catch: cheap, but not everywhere
The GLM Coding Plan subscription offers three tiers: Lite at $12.6/month, Pro at $50.4/month, and Max at $112/month (prices are for annual payment with a 30% discount). At first glance, this is significantly cheaper than a Claude or GPT subscription. However, users complain about weak cloud infrastructure, high token consumption in Max mode, and the model's tendency to get stuck in infinite loops. Many note that the model only truly shines in Max mode, which consumes the quota several times faster.
Verdict: killer or competitor?
There is no clear answer. GLM-5.2 is undoubtedly the best open-source model for programming and autonomous tasks at the moment. In certain long scenarios, it comes very close to Claude Opus 4.8. The open MIT license, the ability to run on your own hardware, and the low entry barrier make it a notable player. However, calling it a "killer" of Claude is an exaggeration. According to most tests, Z.ai itself ranks its model below Opus 4.8. Issues with stability, high token consumption, and weak support do not yet allow it to become a full-fledged replacement for the leader. GLM-5.2 narrows the gap but does not overtake. My professional assessment: it is a powerful tool for developers who value openness and are willing to tolerate shortcomings for the sake of savings, but it is still far from the title of "killer."