Any chance to get support for z.ai GLM models ?

[rbarzic]

Couldn't find any reference to it on gptel github page.

[karthink]

I'm busy with some other features, but a PR for this will be welcome. I don't know anything about z.ai. If it offers an OpenAI-compatible API there will be no code needed, just configuration.

[madanyang]

@karthink it offers both the OpenAI and the Anthropic including thinking mode, and with their generous pro plan it is a good and viable option to choose.

[karthink]

@karthink it offers both the OpenAI and the Anthropic including thinking mode, and with their generous pro plan it is a good and viable option to choose.

I see. Why don't they use an OpenAI-compatible API like all the other providers?

[madanyang]

I think their intention is to make their model available for Claude Code, Cline and others like. So it provides an interface for both Anthropic and OpenAi, one just decides which one is beneficial. I configured it as openai variant and it works, hadn't had time to play with other options.

[karthink]

I think their intention is to make their model available for Claude Code, Cline and others like. So it provides an interface for both Anthropic and OpenAi, one just decides which one is beneficial. I configured it as openai variant and it works, hadn't had time to play with other options.

If it offers an OpenAI/Anthropic-compatible API, then there is no need to add any code to gptel. If you can provide a PR with a README instructions section for z.ai I can merge it.

[rbarzic]

Thank you for the answer @madanyang . Do you mind sharing your setup ? I'll try to set it up when I'll have time but having an setup example may help

[madanyang]

@rbarzic if using the GLM Coding Plan, you need config the exclusive Coding API
https://api.z.ai/api/coding/paas/v4 instead of the Common API https://api.z.ai/api/paas/v4
and append the chat/completions

(gptel-make-openai "CodingPlan"
  :host "api.z.ai"
  :endpoint "/api/coding/paas/v4/chat/completions"
  :stream t
  :key "your-api-key" ;;can be a function that returns the key
  :models '( glm-4.6
             glm-4.5-air
             ))

(gptel-make-openai "Common"
  :host "api.z.ai"
  :endpoint "/api/paas/v4/chat/completions"
  :stream t
  :key "your-api-key" ;;can be a function that returns the key
  :models '( glm-4.6
             glm-4.5
             glm-4.5-air
             glm-4-32b-0414-128k
             ))

Mine is a little bit more complicated so this one follows the gptel-make-openai backend style.

[karthink]

Here's what I'm using:

(let ((meta '( :context-window 128
                 :capabilities (json tool reasoning))))
    (gptel-make-openai "zai"
      :stream t
      :host "api.z.ai"
      :endpoint "/api/paas/v4/chat/completions"
      :key #'gptel-api-key-from-auth-source
      :models
      `(( glm-4.6 :input-cost 0.6 :output-cost 2.2
          :context-window 200
          :capabilities (json tool reasoning))
        (glm-4.5 :input-cost 0.6 :output-cost 2.2 ,@meta)
        (glm-4.5-x :input-cost 2.2 :output-cost 8.9 ,@meta)
        (glm-4.5-air :input-cost 0.2 :output-cost 1.1 ,@meta)
        (glm-4.5-air-x :input-cost 1.1 :output-cost 4.5 ,@meta)
        (glm-4-32b-0414-128k :input-cost 0.1 :output-cost 0.1 ,@meta)
        ( glm-4.5v :input-cost 0.6 :output-cost 1.8
          :capabilities (media)
          :mime-types
          ("image/jpeg" "image/png"
           "application/pdf" "image/gif"))
        glm-4.5-flash)))