Z.ai

@ai-sdk/zai is the official Provider in the Vercel AI SDK for Zhipu AI models.

@ai-sdk/openai-compatible is the common compatibility layer for many OpenAI-compatible model services, and changes to its version may affect multiple provider adapter packages at once.

@ai-sdk/zai depends on @ai-sdk/openai-compatible, indicating that Zhipu GLM's API interface is largely compatible with OpenAI's calling conventions.

Z.ai's choice to integrate through openai-compatible is an efficiency-first strategy, with the cost being that when Z.ai has unique capabilities beyond the OpenAI specification, the compatibility layer may need additional adaptation code.

@ai-sdk/zai, as the Provider implementation for Z.ai, reuses the communication and protocol handling logic of the OpenAI compatibility layer at its foundation.

The API service brand offered externally by Zhipu AI is Z.ai (formerly the BigModel open platform).

GLM-5.3-Flash, from Z.ai, is a multimodal MoE model with a total of 320B (320 billion) parameters, with only 18B parameters activated at any given time.

Z.ai is the brand identity of Zhipu AI for the global market.

The models tested include Google's Gemini 3.7 Flash, Z.ai's GLM-5, Alibaba's Qwen 3 235B A22B Instruct, and OpenAI's gpt-oss-120b, all running on Kaggle.

openai-compatible is a kind of "protocol adapter base" that lets any vendor following the OpenAI Chat Completions interface style (Z.ai, DeepSeek, Together, etc.) integrate with very little code.

After upgrading, it is advisable to regression-test the core paths involving Z.ai model calls, focusing on areas sensitive to provider-utils such as streaming responses and tool calls.

For the vast majority of users, the risk of a patch-version dependency update is low, and it is recommended to keep up with it; however, if the project strictly locks dependency versions, or is troubleshooting anomalies related to Z.ai integration, it is worth comparing the specific diffs of the two dependency packages before upgrading.

The Z.ai adapter follows the AI SDK's overall semantic versioning cadence: incrementing patch numbers correspond to dependency refreshes and internal fixes, and minor version numbers correspond to new capability integration.

Kimi K3 is provided by Moonshot AI, GLM 5.3 is provided by Z.ai, and DeepSeek V4.1 Flash is provided by DeepSeek.ai.

Z.ai claims that GLM-5.3-Flash is close to Claude Opus 4.8 on coding and agent benchmarks and costs one tenth of GLM-5.2's price.

GLM-5.3-Flash is a multimodal MoE model with 320B total parameters and 18B activated per token, under the MIT license, supporting a 1M token context.

IboChat's tech stack includes a single-file Flask application, GLM glm-4.5-air via the Z.ai API, locally running faster-whisper, SQLite + Werkzeug hashed account management, and Docker deployment behind Cloudflare.

IboChat's endpoint name is still /api/gemini because early versions used Google Gemini, and it now proxies to Z.ai; changing the name would break the frontend.

GLM (glm-4.5-air) is Zhipu AI's large language model series, called via the OpenAI-compatible API provided by Z.ai, with interface paths, request bodies, and response formats largely consistent with the OpenAI Chat Completions specification.

In 2026, Kimi, Z.ai (GLM), and Alibaba Cloud Qwen all launched monthly coding plans for developers, with the core logic being to limit usage by 5-hour windows rather than billing by token.