GLM-5.2 vs. Claude: A Real Threat or Just Another Hype?
In the world of artificial intelligence, a wave of discussion is growing around the new GLM-5.2 model from Z.ai. Enthusiasts have already dubbed it the "killer" of Anthropic's Claude, and I decided to figure out how true this claim is.
GLM-5.2 is positioned as a flagship model optimized for long work sessions. The key improvement over its predecessor GLM-5.1 is a stable context window of 1 million tokens, which is five times larger than the previous 200,000. This allows the model to keep vast amounts of code and text in view simultaneously without degrading on ultra-long tasks.
Key features of the model:
- A context of 1 million tokens that does not lose quality during long sessions.
- Two levels of reasoning enhancement: High for a balance between performance and token consumption, and Max for maximum capabilities but with increased resource usage.
- An open MIT license, allowing the model to be run on your own hardware (self-hosting).
- The API price remains at the level of the previous version, which sets it apart from competitors.
The model is already available on HuggingFace and ModelScope, as well as through the GLM Coding Plan subscription and the ZCode desktop agent.
What the benchmarks show
According to Z.ai's own tests, GLM-5.2 demonstrates impressive results. On standard programming tests, the gap with GLM-5.1 is significant: 81.0 vs. 63.5 on Terminal-Bench 2.1 and 62.1 vs. 58.4 on SWE-bench Pro. On Terminal-Bench 2.1, the model came very close to Opus 4.8 (85.0) and surpassed Gemini 3.1 Pro (74.0).
On the FrontierSWE test, where the model handles projects lasting tens of hours, GLM-5.2 lags behind Opus 4.8 by only 1%. On PostTrainBench, it outperforms Opus 4.7 and GPT-5.5, yielding only to Opus 4.8. However, on the ultra-long SWE-Marathon, the gap with Opus 4.8 is already 13%. Nevertheless, GLM-5.2 shows the best results among all open models currently available.
How much does it cost and what's the catch
The GLM Coding Plan subscription offers three tiers. Lite costs $12.6 per month with annual payment, Pro costs $50.4, and Max costs $112. Quota consumption depends on load: a 3x multiplier during peak hours (from 2:00 PM to 6:00 PM Beijing time) and 2x off-peak. Until the end of September, a promotion is active where off-peak usage is charged at 1x.
User opinions
The community reaction is mixed. Strengths: the model is called the strongest open neural network, basic logic is noticeably improved, and in programming it is comparable to GPT-5.5 at a high reasoning level. Users note that GLM-5.2 autonomously handles complex tasks and suggests fixes on its own.
However, critics point to weak cloud infrastructure, high subscription costs, and the model's tendency to get stuck in infinite loops. Many believe the model is tailored exclusively for benchmarks, and in real-world tasks, it's easier to pay for Claude or GPT.
Summary: by benchmarks, we have a flagship, but by real code, it's budget-tier AI.
So, is it a "Claude killer" or not?
There is no clear answer. GLM-5.2 is recognized as the best open model for programming and autonomous tasks. In certain long scenarios, it comes very close to Anthropic's flagship. The open MIT license, self-hosting capability, and low entry barrier make it a notable player.
However, it's bloggers, not benchmarks, who call it a "Claude killer." In most tests, Z.ai itself ranks its model below Opus 4.8. Additionally, users complain about unstable cloud infrastructure, high token consumption in Max mode, and weak support. The new AI narrows the gap with leaders but does not yet surpass them.
My analysis: GLM-5.2 is an important step forward for open models, especially in the context of programming and long tasks. However, calling it a "Claude killer" is premature. It rather demonstrates that Chinese developers are actively catching up with Western giants, but to fully compete, they still have many barriers to overcome related to infrastructure and user experience.