Crypto news

Fierce debates have erupted in the developer and crypto enthusiast communities around the new Chinese model GLM-5.2 from Z.ai. Many have already dubbed it the "Claude killer" from Anthropic, pointing to impressive benchmark results at a price ten times lower than the competitor. However, as a professional AI market analyst, I believe such claims require deeper scrutiny.

What is GLM-5.2 and what is its strength?

GLM-5.2 is a flagship open model optimized for long working sessions. The main innovation compared to version 5.1 is a stable context window of 1 million tokens (up from 200 thousand previously). This allows the model to retain vast amounts of code or text in its "field of view" without loss of quality. Key features:

• 1 million token context: No degradation during ultra-long sessions. The entire codebase fits into a single reasoning cycle.
• Two levels of "effort": High — for a balance of performance and token consumption, Max — for maximum capabilities, but with higher consumption.
• Open MIT license: No regional restrictions, full self-hosting support.
• API price: Remained at the level of the previous version GLM-5.1.

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

What do the benchmarks show?

According to Z.ai's internal tests, GLM-5.2 is recognized as the strongest open model on the market. However, it falls short of Anthropic's flagship — Claude Opus 4.8 — in most cases. On standard programming tests, the gap with GLM-5.1 is noticeable: 81.0 vs 63.5 on Terminal-Bench 2.1 and 62.1 vs 58.4 on SWE-bench Pro. At the same time, on Terminal-Bench 2.1, the result of 81.0 closely approaches Opus 4.8 with its 85.0 and surpasses Gemini 3.1 Pro with 74.0.

Comparison with competitors in maximum reasoning mode:

On long-horizon tasks, the picture is similar. On the FrontierSWE test, where the model manages open technical projects for tens of hours, GLM-5.2 lags behind Opus 4.8 by only 1%, but surpasses GPT-5.5 and the previous version Opus 4.7. On PostTrainBench, which evaluates improving other models through fine-tuning, GLM-5.2 outperforms Opus 4.7 and GPT-5.5, yielding only to Opus 4.8.

Cost and real-world reviews

The GLM Coding Plan subscription is divided into three tiers: Lite — $12.6/month (discounted), Pro — $50.4/month, Max — $112/month. However, within the subscription, quota consumption depends on load: a 3x coefficient during peak hours (14:00-18:00 Beijing time) and 2x off-peak. Until the end of September, a promotion is active where off-peak usage is billed at 1x.

Users note strengths: the model is called the strongest open neural network currently available, the basic logic is noticeably better than version 5.1, and in programming it is comparable to GPT-5.5 at a high reasoning level. However, criticism concerns service and stability: the cloud infrastructure, despite a good mathematical model, is described as extremely weak, and the pricing as expensive. Developers complain about weak support and the model's tendency to get stuck in infinite loops.

Separately, users note that the model only reveals its full potential in Max mode, which consumes many times more tokens than High.

My conclusion as an analyst: GLM-5.2 is a powerful step forward for open models, especially in the context of long sessions and programming. However, calling it a "Claude killer" is premature. According to most tests, Z.ai itself places its model below Opus 4.8. Infrastructure issues and high token consumption in Max mode make it more of a niche tool for enthusiasts willing to sacrifice stability for price. For now, it is not a "killer," but a serious competitor that is closing the gap, though not overtaking the leaders.