Short answer: I couldn’t verify an official model named GLM-4.7V in the documentation reviewed. The available evidence does not establish that GLM-4.7 accepts images. Z.ai’s documentation identifies GLM-4.6V and GLM-4.5V as vision-language models with image input. If you need image understanding, evaluate one of those documented V models rather than assuming “GLM-4.7 V” is a vision variant. (Z.ai’s GLM-4.6V overview)
What does “GLM 4.7 V” refer to?
The name remains unresolved. NVIDIA NeMo AutoModel’s Nightly documentation describes GLM-4.7 as a text-generation model and lists zai-org/GLM-4.7 and zai-org/GLM-4.7-Flash. That page covers NeMo AutoModel; it does not rule out every other product or deployment, but it does not document a GLM-4.7 vision variant. (NVIDIA NeMo AutoModel: GLM-4.7)
There are similarly named models, but they are not interchangeable:
- GLM-4.7-FlashX is a separate name. A SourceForge comparison page discusses it and lists GLM-4.5V-Flash separately as an alternative. That third-party listing does not establish that FlashX is a vision model or that “GLM-4.7 V” refers to either product.
- GLM-4.6V and GLM-4.5V are identified in
Z.ai’s own documentation as multimodal models with image input. Their names and documented inputs are a better basis for vision-model selection. (GLM-4.6V; GLM-4.5V)
Which GLM model should you consider for images?
| If your task is… | Documented candidate | What the official description says |
|---|---|---|
| Analyzing screenshots, images, documents, or mixed visual inputs; workflows involving tool use | GLM-4.6V | Accepts video, image, text, and file inputs. Z.ai describes multimodal tool use and a 128K-token context window. (Z.ai overview) |
| Image or video understanding, visual grounding, GUI tasks, or image-and-text problem solving | GLM-4.5V | Accepts video, image, text, and file inputs; its overview describes visual tasks and a 16K maximum output. (Z.ai overview) |
| Ordinary text generation, with no documented image-input requirement | GLM-4.7 | NVIDIA’s cited NeMo page covers text generation and fine-tuning—not vision or hosted inference. (NVIDIA NeMo AutoModel) |
These are capability-based starting points, not a claim that one model will perform best on your particular images. Test representative inputs for your task before choosing for production.
Is there a production API?
Z.ai documents an API platform and a chat-completion endpoint that supports multimodal inputs, including images. Its general API reference gives the base endpoint as https://api.z.ai/api/paas/v4 and describes Bearer-token authentication; the chat-completion route is /chat/completions. See the API introduction and chat-completion reference. Z.ai’s GLM-4.5V and GLM-4.6V overviews also link to API documentation.
The model-specific integration details are not confirmed here: the exact model ID accepted by the API and the image-content field structure for a GLM vision request. The general API reference establishes the route and multimodal endpoint, but the available information does not establish that either candidate is enabled for a particular account or provide a complete, model-specific image request. Check the API’s model listing and linked request documentation before integrating; don’t treat the general text example as an image example. The documented NeMo fine-tuning workflow for GLM-4.7 is separate and does not prove hosted inference or vision support. (NVIDIA NeMo AutoModel: GLM-4.7)
Do not copy the hypothetical api.example.com/glm47 example found in an unrelated article: it is a placeholder, not a working GLM API instruction.
Cost comparison: Z.ai’s pricing page lists prices in USD per 1 million tokens. It lists GLM-4.6V at $0.30 input and $0.90 output, and GLM-4.5V at $0.60 input and $1.80 output per million tokens. At those listed rates, 1 million uncached input tokens plus 100,000 output tokens would cost $0.39 for GLM-4.6V or $0.78 for GLM-4.5V. This arithmetic uses the published token rates; actual billing depends on how the provider counts a request. Recheck the pricing page before committing, since rates can change.
Practical selection
- Need visual understanding? Start by evaluating GLM-4.6V or GLM-4.5V; don’t select GLM-4.7 on the assumption that “V” means vision.
- Need an API integration?
Z.aidocuments a general multimodal chat-completion route, but the exact vision-model ID and image payload are unconfirmed here. Verify both in the API documentation and confirm the model is available to your account before building around it. - Need self-managed fine-tuning? NVIDIA NeMo AutoModel documents a GLM-4.7 text-generation fine-tuning recipe. That is a separate workflow; it does not prove hosted inference or vision support. (NVIDIA NeMo AutoModel: GLM-4.7)
Bottom line: Treat “GLM-4.7 V” as an unverified model name. For image understanding, compare the documented GLM-4.6V and GLM-4.5V options against your real inputs, API requirements, and budget. Before production integration, confirm the chosen model ID, image format, and account availability in Z.ai’s API documentation.
Documentation and listed pricing reviewed November 10, 2026. Product and API details may change.
Sources
GLM-4.6V — Z.AI Developer Documentation
GLM-4.5V — Z.AI Developer Documentation
Pricing — Z.AI Developer Documentation
API Introduction — Z.AI Developer Documentation