Z.ai: the GLM family of open models
Fine-tune and deploy GLM models from Z.ai with up to 1M-token context on Crusoe infrastructure.
Deploy models quickly
Go live in a few clicks with a Serverless Inference API, or a dedicated self-serve endpoint running on Crusoe-operated infrastructure with no customer data sent to Z.ai.
Millions of tokens without the recomputation tax
Stop paying for prefill recomputation. Crusoe's inference engine is powered by MemoryAlloyâ„¢ technology, a proprietary memory fabric that shares the KV cache across nodes, so repeated long-context calls skip redundant prefill computation.
Start without a commitment
Z.ai models are available for Serverless Fine-Tuning, Serverless Inference, and Self-Serve Deployments in Crusoe Intelligence Foundry. No provisioning, no minimum spend.
Fine-tune and deploy in Crusoe Intelligence Foundry
Frequently
asked questions
Z.ai (known as Zhipu AI in China) is a Beijing-based research lab spun out of Tsinghua University in 2019. In January 2026 it became the first major Chinese LLM developer to go public, listing in Hong Kong. Its GLM models focus on agentic coding, tool use, and very long context. Run them through Crusoe Intelligence Foundry, on Crusoe-operated infrastructure, with no customer data sent to Z.ai. Request a Tailored Deployment for bespoke optimization and a dedicated endpoint.
GLM models are released as open weights, meaning the model files are publicly available to download, run, and fine-tune. Licensing varies by release: some GLM models ship under the MIT license, while others ship under Z.ai’s own GLM model license. Both permit commercial use, modification, and redistribution, and both require you to retain the copyright notice. Z.ai’s license adds conditions for large-scale model-as-a-service operators. Confirm the license on each model card before deploying.
Enterprise teams can run GLM models without sending data to Z.ai by using a provider that hosts the open weights on its own infrastructure. On Crusoe Cloud, GLM models run on Crusoe-operated infrastructure, and no customer data is sent to Z.ai. With open weights, data handling is governed by whoever hosts the model, not by the lab that trained it.
Most teams should start with a lighter GLM variant and move to the flagship model only when answer quality becomes the bottleneck. Lighter variants offer lower latency and cost for high-volume or real-time work, while flagship models handle the hardest reasoning and coding tasks. Test both on a sample of real prompts before committing.
Pricing for GLM models is flexible and depends on how you choose to use them. Serverless Fine-Tuning is billed per 1M training tokens, where training tokens equal dataset tokens multiplied by epochs. Serverless Inference is billed per token, with separate rates for uncached input, cached input, and output tokens, and no minimum commitment. Self-Serve Deployments are billed per GPU hour based on the GPU you choose, giving you predictable costs at production volume. You can get started without long-term commitments in Crusoe Intelligence Foundry, and volume pricing is available as your usage grows. Rates vary by model, so check each model’s details for current pricing. Talk with our team for help finding the right fit for your workload.
Teams that need guaranteed performance can request a dedicated GLM endpoint through a Tailored Deployment on Crusoe Cloud. Crusoe benchmarks the model against the team’s specific workload, then delivers a dedicated endpoint backed by an SLA. This option suits production workloads that need more predictable throughput and latency than shared serverless access provides.
