Z.ai: GLM 4.5 Air

Z.ai: GLM 4.5 Air is a text model from Z.ai, released by the provider on 25 July 2025. It is the lightweight variant of the GLM 4.5 flagship family, built with a Mixture-of-Experts architecture and a more compact parameter count than its larger sibling. Z.ai describes the family as purpose-built for agent-centric applications, and the Air variant carries that focus into a smaller package.

The model works with text input and produces text output. It has a context window of 131,072 tokens and can generate up to 98,304 tokens in a single response, so long inputs and long outputs are both feasible within the same request. Its knowledge cutoff is 31 December 2024. The supported parameters include tools and tool_choice, which means the model can be given callable functions and can decide when to invoke them, a useful property for agent workflows. It also supports a reasoning parameter alongside include_reasoning, so you can control whether reasoning output is requested and whether it is returned.

Those capabilities point to a few practical uses. Agent-style applications are the obvious fit: tool calling plus a large context window lets you feed in instructions, conversation history and tool definitions, then let the model plan and call functions. Long document work is another, since the context window accommodates substantial source material for summarisation, extraction or question answering, and the large output ceiling helps when the result itself is long. Reasoning support suits tasks where you want the model to work through a problem before answering, and the compact size of the Air variant makes it a reasonable choice when you want the GLM 4.5 family behaviour without the weight of the full model.

You call it through api14's OpenAI-compatible API using the model id "z-ai/glm-4.5-air". Because the interface follows the OpenAI format, you can point an existing OpenAI client at api14 and set that id as the model. The request accepts frequency_penalty, presence_penalty, repetition_penalty, temperature, top_p, top_k, seed, stop and max_tokens for controlling sampling and output length, plus tools, tool_choice, reasoning and include_reasoning for tool use and reasoning behaviour. The live price table on this page covers what a request costs.

One limit worth noting is modality: input and output are text only, so if your task involves images or audio, this model is not the fit. The knowledge cutoff of 31 December 2024 also means anything after that date is outside its training data, which matters for time-sensitive questions unless you supply current information in the prompt. If you need a larger model from the same family, the full GLM 4.5 is the alternative; Air is the lighter option.