Model catalogue · Z.ai (Zhipu AI)

GLM

Version via EU inference of open models through the Gosign chain: GLM 5.3 754B , model zai-org/GLM-5.3 (as of 29 September 2026)

EUR 1.625 per million tokens as of 11 October 2026

Vendor: glm-5.3-flash, as of 29 September 2026

An open-weights model that reaches the highest score among open models in the Artificial Analysis index (as of 08/2026); the Flash tier is under the MIT licence, the flagship under the vendor's own licence, and the training focus is on programming and agent tasks rather than German legal register. From day one it is available in every instance through the Gosign chain, as one of the open models Gosign hosts itself in Germany; the contractual chain is closed there, with no US group in the chain. It reasons before every answer and is the most expensive open model on this route. The Open Telekom Cloud also offers the model, under its own standard agreement with its customers. Dedicated operation on reserved hardware and operation on your own hardware fail here not on law but on equipment.

Available through us

Everything section 203 requires is settled, including the provider's confidentiality undertaking, and nothing is stored at the provider, apart from a technical cache of a few minutes for Claude (zero retention). You do not have to clarify anything further.

The operator bills separately - input EUR 1.00 · output EUR 3.50 · cache hits EUR 0.25. The figure above is the blend of the two, weighted three quarters input to one quarter output, as it arises in a firm. Charged at the vendor's list price, without a mark-up.

The providers set these prices; they can change at any time, and our figures may contain errors. The list price in force at the time is authoritative.

What it is good for

  • The Flash tier is under the MIT licence, the freest of the common model licences; the flagship has the vendor's own licence
  • Highest score in the Artificial Analysis index among all open-weights models (as of 08/2026)
  • The Flash tier runs in dedicated operation, the large tier through the EU inference chain

Where it stops

  • The vendor is based in China - irrelevant for the contractual chain, because open weights are operated in Europe, but in need of explanation towards clients
  • The large tier cannot be operated on a single dedicated instance
  • Cut for programming and agent tasks; legal text work is not its training focus
  • No variant of the line with a documented parameter count runs on a single dedicated card - the documented tier with 357 billion parameters needs around 196 GB for the weights alone even in 4-bit quantisation (research date 2026-08-17). The load-bearing route is the open models Gosign hosts itself in Germany

Status through us

The same model, procured differently, is a different legal position. The status above refers to procurement through us; processing takes place in Germany.

Which route carries it, with which instrument and under which condition, we set out in full to the firm before it commissions us - section 203(4) requires that the chain can be checked, not that it is published. The full matrix, with the sources behind every route, can be requested in the German catalogue after confirming your e-mail address, and we go through it in the conversation.

Is GLM the right fit?

We go through your tasks and the route that carries them - and say plainly where it does not.

Arrange a conversation