# GLM 4.5 Air

The efficiency-optimized variant in Z.AI's GLM-4.5 generation, trading a degree of reasoning depth for faster inference and lower per-token cost at high volume.

- **Model ID:** `zai/glm-4.5-air`
- **Type:** chat
- **Providers:** zai
- **Context window:** 128,000
- **Maximum output tokens:** 96,000
- **Pricing:** $0.2/1M input tokens, $1.1/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/glm-4.5-air

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'zai/glm-4.5-air',
  prompt: 'Why is the sky blue?'
})
```

## About

GLM 4.5 Air was released July 28, 2025 as the efficiency-optimized variant in Z.AI's GLM-4.5 generation. Where GLM-4.5 targets maximum capability, GLM 4.5 Air trades a degree of depth for faster inference and lower per-token cost, making it practical for high-throughput production pipelines.

The model retains the core reasoning, coding, and agentic capabilities of the GLM-4.5 family while operating at reduced computational overhead. This positions it for workloads where response latency and cost per request are primary constraints: classification, extraction, summarization, and conversational applications that process high volumes of requests.

GLM 4.5 Air supports the same context window of 128K tokens as the full GLM-4.5 model. Through AI Gateway, it benefits from unified API access, built-in observability, and intelligent provider routing with automatic retries.

## What to consider

GLM 4.5 Air is optimized for speed. For tasks requiring deep multi-step reasoning, the full GLM-4.5 or later models like GLM-5 may produce better results.

At $0.2 input and $1.1 output per million tokens, GLM 4.5 Air is designed for workloads where unit economics matter. Estimate your monthly token volume to compare total cost against heavier alternatives.

GLM 4.5 Air uses the same API interface as GLM-4.5, so switching between the two requires only changing the model identifier.

## When to use

### Best For

- **High-volume production pipelines** where low latency and cost efficiency per request outweigh peak reasoning depth
- **Classification and extraction tasks** that need competent language understanding without extended chain-of-thought overhead
- **Conversational applications** serving many concurrent users where response speed directly affects user experience
- **Summarization workflows** processing large document sets where throughput determines pipeline feasibility
- **Development and prototyping** where fast iteration cycles benefit from quick model responses

### Consider Alternatives When

- Deep reasoning or complex multi-step planning is required. The full GLM-4.5 or GLM-5 provides deeper deliberation capabilities
- Vision or image understanding is needed. GLM-4.5V builds on GLM-4.5-Air with multimodal input support
- Your workload is primarily advanced code generation. GLM-4.6 and GLM-4.7 include targeted coding improvements
- You need the absolute lowest cost for simple tasks and can accept further capability tradeoffs. Evaluate flash-tier models in the GLM lineup

## Best for

- **High-volume production pipelines:** Low latency and cost efficiency per request outweigh peak reasoning depth
- **Classification and extraction tasks:** Competent language understanding without extended chain-of-thought overhead
- **Conversational applications:** Many concurrent users where response speed directly affects user experience
- **Summarization workflows:** Large document sets where throughput determines pipeline feasibility
- **Development and prototyping:** Fast iteration cycles benefit from quick model responses

## Consider alternatives

- **Deep reasoning needed:** The full GLM-4.5 or GLM-5 provides deeper deliberation capabilities for complex multi-step planning
- **Vision or image understanding:** GLM-4.5V builds on GLM-4.5-Air with multimodal input support
- **Advanced code generation:** GLM-4.6 and GLM-4.7 include targeted coding improvements
- **Lowest cost simple tasks:** Evaluate flash-tier models in the GLM lineup when further capability tradeoffs are acceptable

## Frequently asked questions

### How does GLM 4.5 Air compare to the full GLM-4.5?

GLM 4.5 Air is the lightweight variant optimized for faster inference and lower cost. GLM-4.5 provides deeper reasoning capability at higher per-token cost. Both share the same API surface and context window.

### What is GLM 4.5 Air best suited for?

High-volume tasks where speed and cost matter: classification, extraction, summarization, and conversational applications. For deep reasoning tasks, consider the full GLM-4.5 or GLM-5.

### What is the context window for GLM 4.5 Air?

128K tokens, matching the full GLM-4.5 model.

### How do I switch between GLM 4.5 Air and GLM-4.5?

Change the model identifier in your API call. Both models share the same API interface, so no other integration changes are needed.

### How do I authenticate with GLM 4.5 Air through AI Gateway?

AI Gateway provides a unified API key. Configure it in your environment and use the model identifier to route requests. No separate Z.AI account is required, though BYOK is supported.

### Is GLM 4.5 Air suitable for agentic workflows?

Yes, for agent steps that prioritize speed over deep reasoning. For planning-heavy steps, route those to GLM-4.5 or GLM-5 while using GLM 4.5 Air for faster execution steps.

### What is the pricing for GLM 4.5 Air?

Check the pricing panel on this page for today's numbers. AI Gateway tracks rates across every provider that serves GLM 4.5 Air.

## Links

- [Model page](https://vercel.com/ai-gateway/models/glm-4.5-air)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://z.ai/blog/glm-4.5)
- [Provider pricing](https://docs.z.ai/guides/overview/pricing)


If you encounter incorrect information, can’t find what you need, need a capability Vercel doesn’t currently offer, or have any other feedback to report, please do it by following the instructions at https://vercel.com/feedback.md.