The Vercel AI SDK has introduced the Z.AI provider, enabling GLM chat completions, streaming, reasoning, tools, and multimodal inputs. This update also incorporates support for the GLM-5.3-Flash model within the Z.AI provider and the AI Gateway. Additionally, the @ai-sdk/amazon-bedrock@4.0.163 package now preserves complete Amazon Bedrock Converse usage objects in raw usage metadata.
Key Points
- The Z.AI provider has been added to the Vercel AI SDK.
- Z.AI supports GLM chat completions, streaming, reasoning, tools, and multimodal inputs.
- The GLM-5.3-Flash model is now supported by the Z.AI provider and AI Gateway.
- The @ai-sdk/amazon-bedrock@4.0.163 package preserves complete Amazon Bedrock Converse usage objects in raw usage metadata.
- Moonshot streaming tool calls can now be accepted without indices, and choice-level information is preserved.
Context
According to the Vercel AI SDK changelog, these updates are part of ongoing releases to the SDK. The integration of Z.AI and GLM-5.3-Flash expands the range of available models and capabilities for developers using the SDK. The preservation of Amazon Bedrock Converse usage objects provides more detailed metadata for users of that service.
Why It Matters
Developers using the Vercel AI SDK can now access a broader set of models and functionalities, particularly with the addition of the Z.AI provider and GLM-5.3-Flash model. Enhanced usage metadata for Amazon Bedrock Converse offers more transparency and data for cost analysis or performance monitoring.
What To Do
- Review the Vercel AI SDK documentation for details on integrating the Z.AI provider.
- Test GLM-5.3-Flash model capabilities through the Z.AI provider or AI Gateway.
- Examine the raw usage metadata from Amazon Bedrock Converse to leverage the preserved usage objects.
- Note the changes for Moonshot streaming tool calls regarding indices and choice-level preservation.