Use GLM-5.3 with Crush: Setup & API Guide

Connect Crush through the OpenAI-compatible endpoint, select GLM-5.3, then run a small verification task before using it on a repository. Keep credentials in environment variables rather than project files.

Use GLM-5.3 in 4 steps

Step 1 of 4

Try GLM-5.3 in the OneInfer playground without writing code, or create an API key in the Console to call it from your app.

export ONEINFER_API_KEY="your-api-key"

Prerequisites

  • A OneInfer account with API credits
  • A current Crush installation
  • An API key stored outside source control

Configure the connection

  1. Create or copy a OneInfer API key.
  2. Point Crush at the OpenAI-compatible endpoint.
  3. Choose GLM-5.3 as the model.
  4. Restart the tool so it reloads environment settings.
Example
export OPENAI_BASE_URL="https://api.oneinfer.ai/v1/ula"
export OPENAI_API_KEY="$ONEINFER_API_KEY"
# model: glm-5.3

Verify it worked

Ask for a short, deterministic task such as explaining one local function. Confirm the response uses the intended model before sending a large repository or document.

Tool-specific operating tips

  • Start with a narrow file set before repository-wide work.
  • Require a test plan before edits.
  • Review generated commands before execution.
  • Use higher effort only for tasks that benefit from deeper planning.

Ready to test the workflow?

Troubleshooting

SymptomCheck
401 or 403Confirm the key and authentication header.
Model not foundUse the exact model identifier exposed by the OneInfer catalog.
Unexpected modelRemove cached/default tool configuration and select GLM-5.3 explicitly.
High usageReduce supplied context and cap output for routine tasks.

Credential safety

Never paste an API key into a prompt, commit it to a repository, or expose it in browser-delivered code.

Frequently asked questions

Can I try GLM-5.3 before integrating it?

Use the OneInfer GLM-5.3 launcher to open a prepared prompt in the authenticated playground. Availability is checked against the current model catalog.

How should I treat benchmark claims?

Check the provenance label and harness version. Vendor-reported and independently verified results are deliberately shown as different evidence classes.

Put GLM-5.3 to work

Fund a controlled evaluation, start with a prepared prompt, and measure quality and cost on your own workload.