GLM-5.2: Z.ai's Open-Weights Model Outperforms GPT-5.5 on Coding Benchmarks

What Is GLM-5.2

GLM-5.2 is a 753-billion-parameter open-weights AI model released by Z.ai, the Chinese AI startup formerly known as Zhipu AI. The model launched on June 16 and is distributed under an MIT open-source license through Hugging Face, making it freely available for developers to download, modify, and deploy.

What sets GLM-5.2 apart from many other open releases is that it doesn't just compete with closed, proprietary systems — on several coding benchmarks, it actually beats OpenAI's GPT-5.5, while running at a fraction of the operating cost.

Benchmark Performance: How GLM-5.2 Stacks Up

Coding Benchmarks

GLM-5.2's strongest results show up on long-horizon coding tasks, the kind that require a model to reason through multi-step problems rather than produce a single quick answer.

On SWE-bench Pro, GLM-5.2 posted a score of 62.1, ahead of GPT-5.5's 58.6. On Terminal-Bench 2.1, it became the first open-weights model to break the 80% mark, scoring 81.0. That said, it still falls short of GPT-5.5 (84.0) and Claude Opus 4.8 (85.0) on that particular benchmark, showing that while GLM-5.2 has closed much of the gap with the top closed-source models, it hasn't fully caught up across every test.

Reasoning Performance

Beyond coding, GLM-5.2 also turned in a strong result on AIME 2026, a math reasoning benchmark, scoring 99.2 — ahead of both GPT-5.5 and Claude Opus 4.8 on that particular test.

Pricing and Cost Efficiency

API Pricing Breakdown

Z.ai prices GLM-5.2 access at $1.40 per million input tokens and $4.40 per million output tokens. Cached input tokens are considerably cheaper, coming in at around $0.26 per million.

Cost Compared to GPT-5.5

According to VentureBeat's reporting, this pricing structure puts GLM-5.2 at roughly one-sixth the cost of GPT-5.5 for comparable coding workloads — a significant gap given how close the two models are on raw benchmark performance.

Context Window and Technical Capabilities

GLM-5.2 ships with a one-million-token context window, giving it the capacity to process very large amounts of text, code, or documentation in a single pass. It also supports output of up to 131,072 tokens per response, which matters for tasks like generating long-form code or detailed multi-part answers in one go.

Availability and Access Points

Z.ai rolled out access to GLM-5.2 in two stages. Paid subscribers to Z.ai's coding plan got access first, starting June 13. The full open-weights release followed three days later, on June 16.

Today, the model is available through several channels: Z.ai's own API, Cloudflare Workers AI, and more than 20 third-party coding environments, giving developers a range of ways to integrate it into existing workflows.

GLM-5.2 in the Broader Open-Source Landscape

Fourth Release in the GLM-5 Line

GLM-5.2 is the fourth model in the GLM-5 series since February 2026, following GLM-5, GLM-5-Turbo, and GLM-5.1. That release cadence — four models in just a few months — reflects how quickly Z.ai has been iterating.

Framed as a Response to U.S. AI Restrictions

Z.ai founder Jie Tang positioned the release partly as a reaction to recent restrictions placed on U.S.-based AI services, situating GLM-5.2 within a broader pattern of increasingly competitive open-source models coming out of Chinese AI labs.