Chinese AI lab Z.ai released GLM-5.3 on Thursday through its GLM Coding Plan and ZCode.

The company describes the model as the "most capable open-weights model for coding." GLM-5.3 is a 743-billion-parameter model built by scaling post-training on the GLM-5.2 base. "Open weights" means a lab publishes the trained model files so developers can download and run them on their own hardware, in contrast to closed models such as Anthropic's Claude or OpenAI's GPT series, which are reachable only through a vendor's API.

Z.ai said the model is available now through the GLM Coding Plan subscription and ZCode, with API access and downloadable weights to follow after a safety review.

"Scaling post-training is all we did for GLM-5.3," the company wrote in its launch post. "With GLM-5.2 we built the stack... Over the past month we kept scaling on this stack: more environments, more diverse tasks, and more compute spent training on them."

The team said it prioritized token efficiency rather than raw dominance. GLM-5.3 has 743 billion parameters and uses significantly fewer tokens per task than its predecessor. Parameters are the number of variables a model uses while processing information, while tokens are the basic units of information a model can consume or generate. Token economy also carries a direct commercial effect: because APIs are billed per token, producing fewer output tokens per task lowers the cost of running the model.

Z.ai said GLM-5.3 scores 34.5% on its in-house Z.ai Code Bench at Max effort while using about 75,000 output tokens per task, compared with GLM-5.2's 23.4% at 96,000 tokens. Against closed models, the blog says it beats Claude Opus 4.8 on token economy but "remains behind Claude Fable 5, which reaches 39.5% at Max effort." The comparisons are drawn from Z.ai's own launch post, so they are vendor-reported figures rather than results from independent evaluation.

On coding benchmarks, GLM-5.3 performs well and outperforms fellow Chinese model Kimi K3 on the most relevant tests.

On Terminal Bench 3.0, a test of autonomous shell and tool use in real Linux environments, GLM-5.3 scored 28.3, slightly below closed models Fable 5 at 33.7 and GPT-5.6 Sol at 34.6. On DeepSWE v1.1, a benchmark for fixing real GitHub issues end to end, open rival Kimi K3 scored 67.5 and Fable 5 scored 69.7, both ahead of GLM-5.3's 66.9.

Taken together, the results show GLM-5.3 ahead of its predecessor and some open-weight peers, while closed U.S. models still lead the main coding benchmarks.

The cybersecurity results showed another notable gain. GLM-5.3 led CyberGym, a benchmark that scores models on real-world software exploitation tasks, at 84.5% and more than doubled GLM-5.2 on exploitation benchmarks. Z.ai said the model flagged 2,436 vulnerabilities across 269 open-source projects, including 1,097 medium-to-high severity issues.

"GLM-5.3 takes agentic coding to the next level, delivering a dramatic improvement over GLM-5.2 while achieving better results with fewer output tokens," Z.ai posted on X. "GLM-5.3 is available now through GLM Coding Plan and ZCode. API access and open weights will be released in stages following rigorous safety evaluations."

"GLM-5.3 takes agentic coding to the next level, delivering a dramatic improvement over GLM-5.2 while achieving better results with fewer output tokens. pic.twitter.com/KGc6ZR7GHv — Z.ai (@Zai_org) August 14, 2026"

On pricing, Z.ai is positioning the model as a lower-cost alternative to U.S. frontier systems. Its GLM Coding Plan uses a points quota, with off-peak calls costing half, while Zhipu's API is priced at roughly one-tenth of U.S. frontier per-token rates. GLM-5.2's official pricing was $1.40 in and $4.40 out per million tokens. That compares with GPT-5.3-Codex at $1.75 in and $14 out, and Claude Opus 4.8 near the top of Anthropic's pricing tiers.

Z.ai is a Beijing lab on the U.S. Entity List, meaning American firms cannot export controlled technology to it. Z.ai is the brand of Zhipu AI, a company founded in 2019 as a spinout from Tsinghua University research, which the U.S. Commerce Department added to the Entity List in January 2025. Even so, GLM remains a widely used model, and Chinese open-weight models already lead American ones on OpenRouter token usage; OpenRouter is an API aggregator that routes requests across many models and publishes usage share by model.

According to the launch post, GLM-5.3 weights are expected to be released publicly in about two weeks. The open-weights designation therefore applies to the upcoming release, not to what is downloadable today. Once the weights and API access are available, outside developers will be able to run the model on their own infrastructure and test the company's benchmark claims directly.