# GLM 5 Turbo

A speed-optimized GLM-5 variant that keeps selectable thinking modes while cutting latency and cost for agentic pipelines and high-throughput production traffic.

- **Model ID:** `zai/glm-5-turbo`
- **Type:** chat
- **Providers:** zai
- **Context window:** 202,800
- **Maximum output tokens:** 131,100
- **Pricing:** $1.2/1M input tokens, $4/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/glm-5-turbo

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

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

const result = streamText({
  model: 'zai/glm-5-turbo',
  prompt: 'Why is the sky blue?'
})
```

## About

GLM 5 Turbo was released March 15, 2026 as the speed-optimized variant in Z.AI's GLM-5 generation. The GLM-5 generation introduced selectable thinking modes so you can dial reasoning depth per request, and GLM 5 Turbo makes that capability affordable at production scale.

Agentic pipelines benefit the most. Many pipeline steps don't require the full GLM-5's deliberation depth, but they do benefit from the structured thinking modes when problems get harder. GLM 5 Turbo lets you route routine steps to a lightweight thinking mode for fast responses, then escalate harder steps to a deeper mode, all within the same model and API call format.

The turbo variant also inherits GLM-5's improved long-range planning and agentic coding capabilities. Combined with the lower per-token cost and faster throughput, this makes it practical to run multi-step agent workflows that would be prohibitively expensive at full GLM-5 pricing. Through AI Gateway, GLM 5 Turbo shares the same API surface as GLM-5.

## What to consider

GLM 5 Turbo supports multiple thinking modes like GLM-5. Match the mode to the task: lightweight modes for extraction and classification, deeper modes for multi-step reasoning. Even deeper modes run faster than the equivalent on the full GLM-5.

For the most complex reasoning chains, the full GLM-5 will produce higher-quality results. Benchmark both on your hardest tasks to quantify the difference.

Switching between GLM-5 and GLM 5 Turbo requires only changing the model identifier. No integration changes needed.

## When to use

### Best For

- **High-volume agentic pipelines** where most steps need GLM-5-class capability at lower latency and cost
- **Structured data extraction** from documents where speed matters as much as accuracy
- **Real-time coding assistance** where fast responses improve developer productivity without sacrificing agentic capabilities
- **Production deployments at scale** where per-token cost directly impacts margins
- **Multi-step workflows** mixing fast execution steps on GLM 5 Turbo with complex reasoning steps on the full GLM-5

### Consider Alternatives When

- Maximum reasoning depth is required on every request. The full GLM-5 provides the deepest deliberation in the generation
- Vision or multimodal input is needed. GLM-5V-Turbo adds image understanding to the turbo tier
- Your workload is primarily frontend code generation. GLM-4.7 offers targeted frontend improvements at lower cost
- You need the absolute fastest inference with minimal capability. GLM-4.7-FlashX provides the lowest latency option

## Best for

- **High-volume agentic pipelines:** Most steps need GLM-5-class capability at lower latency and cost
- **Structured data extraction:** Documents where speed matters as much as accuracy
- **Real-time coding assistance:** Fast responses improve developer productivity without sacrificing agentic capabilities
- **Production deployments at scale:** Per-token cost directly impacts margins
- **Multi-step workflows:** Fast execution steps on GLM 5 Turbo pair with complex reasoning steps on the full GLM-5

## Consider alternatives

- **Maximum reasoning depth:** The full GLM-5 provides the deepest deliberation in the generation on every request
- **Vision or multimodal input:** GLM-5V-Turbo adds image understanding to the turbo tier
- **Frontend code focus:** GLM-4.7 offers targeted frontend improvements at lower cost
- **Absolute fastest inference:** GLM-4.7-FlashX provides the lowest latency option when minimal capability is acceptable

## Frequently asked questions

### How does GLM 5 Turbo compare to the full GLM-5?

GLM 5 Turbo shares GLM-5's core capabilities, including multiple thinking modes and enhanced agentic coding. It's optimized for faster inference at lower cost, with some reduction in peak reasoning depth on the most complex tasks.

### Does GLM 5 Turbo support multiple thinking modes?

Yes. It retains GLM-5's multiple thinking modes, letting you select the reasoning depth per request. All modes run faster than their equivalents on the full GLM-5.

### What is the context window for GLM 5 Turbo?

202.8K tokens.

### Can I switch between GLM-5 and GLM 5 Turbo easily?

Yes. Both share the same API surface. Change the model identifier to switch between them without any other integration changes.

### How do I authenticate with GLM 5 Turbo through AI Gateway?

AI Gateway provides a unified API key. No separate Z.AI account is needed. Use the model identifier to route requests. BYOK is also supported for direct provider access.

### Is GLM 5 Turbo good for agentic coding?

Yes. It inherits GLM-5's improvements in autonomous tool use, code planning, and multi-step iteration. The faster inference makes it practical for agent loops where speed compounds across many steps.

### What is the pricing for GLM 5 Turbo?

See the pricing section on this page for today's rates. AI Gateway exposes each provider's pricing for GLM 5 Turbo.

## Links

- [Model page](https://vercel.com/ai-gateway/models/glm-5-turbo)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://docs.z.ai/guides/llm/glm-5-turbo)
- [Provider pricing](https://docs.z.ai/guides/overview/pricing)