glm-5
Configured
Thinking-mode reasoning and chat gateway.
Configuration record
GLM 5 and GLM 4 thinking-mode models.
`glm-5` and `glm-4` use `reasoning_content` playback through the OpenAI Chat Completions channel. Neither is exposed in the main picker.
Floatboat · GLM
Configured below the picker
GLM
What do you want to finish?
Reason through a Chinese document
Not exposed in the main client picker
- Model IDs
- glm-5 · glm-4
- Protocol
- OpenAI Chat Completions
- Reasoning field
- reasoning_content
- Availability
- Historical configuration
A reasoning adapter prepared to preserve thinking-mode output.
The GLM adapter supports reasoning playback through the compatible chat gateway.
Thinking-mode playback
Reasoning
The adapter preserves `reasoning_content` in the response flow.
Chat Completions compatible
Transport
GLM is connected through the configured chat channel.
GLM 5 and GLM 4
Versions
Both model IDs are recognized by the runtime profile.
Protocol ready, picker hidden
Boundary
The adapter status does not imply direct client access.
Two configured GLM model IDs.
The GLM records identify GLM 5 and GLM 4 profiles through the configured adapter and chat gateway. Read each profile with its protocol assumptions, and verify the current entry point separately rather than inferring client availability from configuration alone.
glm-4
Configured
Thinking-mode reasoning and chat gateway.
Configuration record
Use Cases for GLM Models
Reasoning playback and chat-gateway compatibility are integration questions, not just model-selection choices. Examine field mappings, streaming behavior, and access prerequisites separately, with unverified routes kept distinct from configurations already recorded for GLM.
Reasoning playback
Reasoning playback can be checked with compatible response samples and the thinking-mode settings used to produce them. Compare how reasoning_content is mapped and replayed, looking for missing or duplicated segments. Record the field-level differences and correction suggestions before relying on playback in a longer workflow.
Gateway compatibility
Adding a model through an existing chat gateway requires more than changing its name. Compare the request format, model identifier, streaming events, error responses, and response-consumption logic. Document the mapping and run compatibility checks so downstream code can handle the route's actual contract.
Workflow compatibility review
A workflow that depends on reasoning replay needs its model identity, route, and access conditions checked together. Compare those requirements with the current environment and mark every unverified step. The assessment should make clear which compatibility checks have passed and which prerequisites still prevent execution.
Check interrupted and failed responses
A gateway integration also needs to handle partial output and error responses. Compare the consumer's expected events with available response samples, including how reasoning fields and final content are separated. Record where state becomes incomplete or duplicated, and test recovery behavior without treating a configured GLM adapter as proof that the entire client path works.
Protocol support and current access.
Model identity, thinking-mode handling, and gateway behavior all affect a GLM integration. Compare these recorded capabilities as parts of one request-and-response path, with reasoning-field replay checked separately from whether the current client exposes a usable model entry.
GLM 5 profile
`glm-5` is recognized in the configured runtime matrix.
GLM 4 profile
`glm-4` is recognized alongside GLM 5.
Thinking-mode handling
The adapter supports replay of the `reasoning_content` field.
Chat gateway path
The models use an OpenAI Chat Completions-compatible channel.
Check GLM protocol support and available access paths.
Useful for evaluating
- You need the exact GLM adapter behavior.
- Thinking-mode playback is part of the integration review.
- You are distinguishing protocol support from product availability.
Use a live picker model when
- You need to run the task in Floatboat now.
- A currently selectable reasoning model meets the need.
- You need a product workflow rather than integration status.
Explore Other Model Families
Compare model families by the kind of work they support. Availability differs by model and entry point.
GLM FAQ
What are GLM 5 and GLM 4?
GLM 5 and GLM 4 are Zhipu model profiles configured for chat and thinking-mode reasoning through a compatible gateway.
What are the GLM model IDs?
The configured model IDs are
glm-5 and glm-4.What does GLM thinking mode support?
The adapter supports playback of the
reasoning_content field so reasoning output can remain available in the response flow.Which protocol does GLM use?
The configured path uses an OpenAI Chat Completions-compatible channel.
What is the current GLM support status?
Protocol support is configured for GLM 5 and GLM 4, but neither had a main-picker entry in the October 9, 2026 record; check the current client menu.
Does reasoning output remove the need to test GLM tool calls?
No. Evaluate the actual tool arguments, returned evidence, and final state of the task. A plausible explanation does not establish that a tool used the intended file, respected permissions, or completed the requested change.
Define the reasoning fields your workflow must preserve
Use a compatible response sample to specify which reasoning fields must survive mapping and replay. Check the current model route before testing for missing or duplicated content.