Crypto news

Chinese company Z.ai has unveiled its new flagship model GLM-5.2, sparking heated debates in the crypto and AI communities. Many are calling it a "Claude killer," especially given that the API cost of the new model is ten times lower than that of its competitor. I analyzed the technical specifications, benchmarks, and user feedback to understand how justified these claims are.

What is GLM-5.2 and its key features

GLM-5.2 is a model optimized for long working sessions. Its main advantage over its predecessor GLM-5.1 is a stable context window of 1 million tokens, five times larger than the previous version. This allows the model to retain vast amounts of code and text in its "field of view" without quality degradation.

Key characteristics:

• 1 million token context without performance loss on ultra-long tasks.
• Two reasoning levels: High for a balance of speed and cost, Max for maximum performance but with higher token consumption.
• Open MIT license with no regional restrictions, allowing the model to be run on your own hardware (self-hosting).
• API price remains at the level of GLM-5.1, making it extremely attractive.

The model is available on HuggingFace and ModelScope, as well as through the GLM Coding Plan subscription and the ZCode desktop agent.

Benchmarks: Where GLM-5.2 excels and where it falls short

According to Z.ai's own tests, GLM-5.2 is recognized as the strongest open model on the market. However, it still falls short of Anthropic's flagship, Claude Opus 4.8. The gap is particularly noticeable on complex tasks such as NL2Repo (generating an entire project from a description) and DeepSWE, where Opus 4.8 confidently leads.

Nevertheless, on the Terminal-Bench 2.1 test, GLM-5.2 (81.0) closely approaches Opus 4.8 (85.0) and surpasses Gemini 3.1 Pro (74.0). On SWE-bench Pro and ProgramBench, it also shows excellent results, outperforming GPT-5.5 and Gemini.

Long-horizon tasks deserve special attention: on the FrontierSWE test, where the model works on projects for dozens of hours, GLM-5.2 lags behind Opus 4.8 by only 1%. This is a significant achievement.

Cost and subscription: Cheap, but not without nuances

The GLM Coding Plan subscription includes three tiers: Lite ($12.6/month), Pro ($50.4/month), and Max ($112/month), all with a 30% discount for annual payment. Quota consumption depends on load: a 3x multiplier during peak hours (14:00–18:00 Beijing time) and 2x off-peak. Until the end of September, a promotion applies where off-peak usage is charged at 1x.

What users are saying: Enthusiasm and criticism

Opinions are divided. Strengths, according to reviews:

• The model is called the strongest open neural network currently available.
• Logic and programming are at a high level, comparable to GPT-5.5 in Max mode.
• It autonomously performs complex tasks and suggests fixes on its own.
• Although it is described as slow and expensive in Max mode, it is extremely persistent in achieving its goal.

Criticism primarily concerns service and stability:

• The cloud infrastructure is weak; users complain about high costs and poor support.
• The model tends to get stuck in infinite loops and ignore commands.
• Many believe it is tailored exclusively for benchmarks, not for real-world code.

Conclusion: A "Claude killer" or not?

There is no clear answer. GLM-5.2 is the best open model for programming and autonomous tasks today. In certain scenarios, it comes very close to Anthropic's flagship. The open MIT license, self-hosting capability, and low price make it a notable player.

However, calling it a "Claude killer" is more of a hype from bloggers than an objective assessment. In most tests, Z.ai itself ranks its model below Opus 4.8. Users complain about unstable infrastructure, high token consumption in Max mode, and poor support.

My analysis: GLM-5.2 narrows the gap with the leaders but does not yet surpass them. It is a powerful tool for developers who value open-source and are willing to tolerate infrastructure shortcomings. For the mass market, Claude and GPT remain more reliable choices. But the fact that Chinese models are catching up to Western giants so quickly is a signal that cannot be ignored.