> ## Content Index
> Fetch the complete content index at: https://www.tao.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Chutes Launches GLM 5.2 Fast With 1M-Token Context and OpenAI-Compatible Access
- URL: https://www.tao.media/chutes-launches-glm-5-2-fast-with-1m-token-context-and-openai-compatible-access/
- Published: 2026-11-10T16:34:38.000Z
- Updated: 2026-11-10T16:34:38.000Z
- Description: The Bittensor inference platform adds a GLM option for latency-sensitive chat and agents, with separate rates for input, output, and cached tokens.
- Author: Tristan Hillerich
- Tags: Chutes, Subnets, AI, News

Chutes has [launched GLM 5.2 Fast](https://x.com/chutes%5Fai/status/2119807595112182224?ref=tao.media), adding a model for latency-sensitive chat and AI agents to its hosted inference service. The deployment has a 1 million-token context window and up to 128,000 tokens of output, with standard pricing of $2.45 per million input tokens and $7.70 per million output tokens.

Developed by [Z.ai](http://z.ai/?ref=tao.media), the model is available through Chutes' OpenAI-compatible API. Existing Chutes users can access it with the same API key they use for other models on the platform.

Chutes runs on [Bittensor Subnet 64](https://chutes.ai/llms.txt?ref=tao.media), where independent GPU operators supply compute for its decentralized, serverless inference service. The addition gives developers another hosted model option without requiring them to deploy the model weights or manage the underlying GPU infrastructure.

## Long Context for Chat and Agent Workflows

The 1 million-token context window gives applications room to include extensive documents, conversation histories, or code alongside a request. Tokens are the chunks of text a model processes, and the context window sets the working space available for a request and its response.

The separate 128,000-token maximum output controls how much the model can generate in one response. Applications can set a lower output limit to suit the task, whether they need a brief chat reply or a longer generated document.

Chutes is targeting chat and agent workloads where response time matters. A model call for an agent may be one step in a longer sequence of tool use, code execution, and follow-up requests, so response delays can accumulate across a task. Z.ai designed GLM 5.2 for long-horizon tasks that require sustained work over multiple steps, Chutes said.

The [Chutes model listing](https://chutes.ai/app/chute/chutes-zai-org-glm-5-2-fast?ref=tao.media) also links to open weights, a separate route to the model beyond hosted API access.

## Coding Scores From the GLM 5.2 Release

![GLM 5.2 Fast at a glance (settings and benchmarks)](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HWsRxjPXAAAZKv1.jpg)

[GLM 5.2 Fast at a glance (settings and benchmarks) | Chutes (X)](https://x.com/i/status/2119807597200974049?ref=tao.media)

Chutes shared [Z.ai's published GLM 5.2 results](https://x.com/i/status/2119807597200974049?ref=tao.media) alongside the launch. GLM 5.2 scored 81.0 on Terminal-Bench 2.1, which evaluates agentic work in a terminal, compared with 85.0 for Opus 4.8.

On SWE-bench Pro, a software-engineering benchmark, GLM 5.2 scored 62.1, compared with 58.4 for GLM 5.1\. That represents a 3.7-point increase over the earlier model.

These are Z.ai's figures for GLM 5.2, relayed by Chutes, rather than Chutes' own testing of the Fast deployment. They provide coding-performance context for the model family, not a measurement of the hosted service's response speed.

## API Access and Token Pricing

Developers can use an OpenAI-compatible client pointed at the [Chutes API base URL](https://llm.chutes.ai/v1?ref=tao.media), authenticate with a Chutes API key. The key supplied as a Bearer token.

A minimal Python request using the OpenAI SDK looks like this:

```py
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://llm.chutes.ai/v1",
    api_key=os.environ["CHUTES_API_KEY"],
)

response = client.chat.completions.create(
    model="zai-org/GLM-5.2-Fast",
    messages=[{"role": "user", "content": "Explain how token pricing works."}],
    max_tokens=1024,
)

print(response.choices[0].message.content)
```

Chutes lists three token rates:

| Token Category | Price per 1 Million Tokens |
| -------------- | -------------------------- |
| Input          | $2.45                      |
| Output         | $7.70                      |
| Cached input   | $0.49                      |

The cached-input rate is 80% below the standard input rate. For workloads that reuse cached material, such as a common document or instruction prefix, that lower rate reduces the cost of processing repeated input.

Output is billed separately from input. A request consuming 100,000 uncached input tokens and generating 10,000 output tokens would cost $0.322 at the listed rates: $0.245 for input and $0.077 for output.