GLM-5.2 on your own machine
Z.ai's open-source SOTA: 753B, 1M context, MIT license
- Video memory
- 800 GB
- Setup
- ~15 min
- Access
- port 8000
- Billing
- hourly, in BRL
GLM-5.2 is one of the very-long-context open models, which is why it needs a multi-GPU machine. This template brings up vLLM with sharding already configured.
Serve GLM-5.2 (Z.ai) via vLLM with an OpenAI-compatible API. 753B MoE (~40B active), 1M context, strong at agents and code. NOTE: the full model needs multiple high-VRAM H100/H200 GPUs or a quantized variant — tune tensor-parallel and the model via env vars.
What it is for
- Analyse a very long document in a single pass
- Coding agents with wide repository context
- An open alternative to closed frontier models
- Prompts and answers that never touch a third-party API
How to deploy
- Create your account and add balance (card or Pix, no subscription).
- In the console, pick the GLM-5.2 (vLLM) template and a machine — the console hides the ones that do not meet the requirement.
- In about 15 minutes the setup finishes and the access address shows up in the panel, on port 8000.
Done? Just destroy the machine and billing stops with it. No contract, no minimum commitment.
FAQ
How much video memory do I need?
Around 320 GB, meaning a block of several 80–141 GB cards. Check availability in the console before reserving.
How long does it take to come up?
Downloading the weights is the slow part: allow several minutes before the server answers.
Is it worth it against a token API?
Only under heavy, steady use. At low volume the multi-GPU machine does not pay for itself.