Versely
    Guides

    GLM 5.3 is a planner, not a renderer

    Pin z-ai/glm-5.3 behind the Versely agent. Text-only, 1M context, reasoning always on. Name Kling or Seedance in the same brief. Do not attach a still.

    Versely Team••4 min read

    Pin z-ai/glm-5.3 behind the Versely agent. It plans and calls tools. It does not draw the clip. Name Kling 3 Turbo or Seedance 2.5 in the same brief. Do not attach a still — the GLM-5 family is text-only, and the image part 404s.

    GLM-5.3 is Z.ai’s large-scale reasoning model: text in, text out, 1,048,576-token context, up to 131,072 completion tokens, function calling. OpenRouter list price is $1.40 per million input tokens and $4.40 per million output, cache read $0.26. Released 18 August 2026. Reasoning is always on and cannot be disabled; efforts are low, high, and max (max is the default). That is a planner profile. It is not a row on /models.

    Two catalogs

    Picking the chat model is the rule. Chat models live on GET /api/v1/agentic/chat-models. Generation models live on the public catalog. If chat_model is not in the built-in map but looks like org/model, it is a custom OpenRouter brain. z-ai/glm-5.3 is that shape.

    The current documented default is runpod/kimi-k2.7-code. Older Versely posts named z-ai/glm-5.1 as a cheap orchestrator. 5.3 is a different slug. Saving “GLM” as a preference does not upgrade you. Send the id.

    Chat tokens and generations bill on two meters. Changing to 5.3 changes the reasoner line. It does not change what a five-second Kling job costs once the renderer is named. There is no free chat allowance.

    Text-only is the product constraint

    From the built-in registry, the GLM 5 family is text-only. MiniMax M2.7, Qwen3 Max Thinking, the MiMo line, DeepSeek V4 Pro, and the RunPod Kimi ids share that constraint. Send an image to those and the provider 404s on the image part.

    Practical rule:

    • Brief is already words: GLM 5.3 is a fair pin for a long tool-calling loop.
    • Brief is a pack shot, a screenshot, a frame: pick a vision brain (Grok 4.6, Grok 4.3, Gemini 3.1 Pro, Claude, Qwen3.5 397B) or put the visual in text and skip the attachment.
    • Video or audio attachments are pre-processed through Gemini for non-Gemini brains. You do not have to pick Gemini just to describe a clip. You do have to pick vision if the image is the brief.

    Do not “fix” the 404 by asking GLM to imagine the product. That is a generate. Upload the still to a vision chat model, or generate from a locked file with image-to-video.

    Pin it, then name the renderer

    On POST /api/v1/agentic/chat (and /chat/stream) send:

    • chat_model: z-ai/glm-5.3
    • chat_model_explicit: true when this is a choice, not a leftover default

    On the OpenRouter path, fast-eligible sub-agents (generation, slideshow, social, web editor) may run on Claude Haiku 4.5 unless that flag is set. Workflow scene-directing and movie planning keep the brain you picked. The default RunPod Kimi is not swapped for Haiku.

    In the agent UI, pick or save the OpenRouter id. A lasting preference can store “plan with GLM 5.3.” It still does not pin Kling.

    Prompt shape:

    Plan with GLM 5.3. Do not change the video model. Generate with Kling 3 Turbo, 5s, 9:16, locked-off bottle. Call the schema tool if you need allowed durations. Do not substitute a different renderer.

    If you omit the second sentence, a newly long-horizon brain may “help” by routing to a different family. You will blame GLM for a look it did not render.

    What 5.3 is not

    It is not Veo. It is not Seedance. It is not Grok Imagine. It is not captions, publishing, or a second Versely subscription. It is not GLM-5V Turbo — that is a different Z.ai id if you need vision from that lab.

    It is also not a Claude Code plugin. The Versely skills install is npx skills add AI-XLabs-Innovation/versely-skills. That pulls generate / slideshow / movie / ugc / music / social / analytics / pipeline into the host agent. GLM 5.3 is the in-app planner when you are already in Versely chat.

    FAQ

    Is GLM 5.3 in the built-in picker by default?

    Treat /chat-models as live. As of the two-catalog write-up, the default is Kimi K2.7 Code. 5.3 is a custom OpenRouter id you send or save.

    Why did my photo attachment fail?

    GLM 5.3 is text-only. Attachments that are images 404. Use Grok 4.6, Gemini 3.1 Pro, or Claude for stills-as-brief.

    Does always-on reasoning mean I pay more?

    OpenRouter bills input and output tokens at the listed rates. Reasoning cannot be turned off on this id. For templated one-liners, pick a cheaper or no-reasoning brain. For a multi-step campaign plan, 5.3 is the point.

    Will this replace Kling in my workflow?

    No. GLM decides the tool calls. Kling (or Seedance, Veo, Grok Imagine Video) decides the file. Pin both.