What is the GLM-5.2 context window?

Z.AI documents a 1M-token context window and a 128K maximum output for GLM-5.2, and describes the long context as lossless rather than merely available. Reasoning tokens count toward the output limit, so a long thinking pass shares its allowance with the visible answer.

Can I turn off thinking on GLM-5.2?

Yes. Thinking is enabled by default and can be disabled explicitly, which is not true of every model in this class. Left enabled, GLM-5.2 judges for itself whether a given request needs reasoning — Z.AI contrasts this with the previous generation, which reasoned unconditionally.

Does GLM-5.2 accept temperature and top_p?

Yes — both, and this is a real difference from several open-weight peers that pin them. GLM-5.2 takes temperature from 0.0 to 1.0 with a default of 1.0, so mind the ceiling if your code sends higher values, and top_p from 0.01 to 1.0 with a default of 0.95. Z.AI recommends tuning one of the two rather than both.

Can GLM-5.2 read images?

No. GLM-5.2 takes text in and returns text out; vision is a separate model in Z.AI's lineup. If your workflow needs a model to look at a screenshot or a scanned document, GLM-5.2 is the wrong choice for that step even though it is a strong choice for the code around it.

Which endpoint and model string does GLM-5.2 use on reAPI?

An OpenAI-compatible `/v1/chat/completions` endpoint on api.reapi.ai, with `glm-5.2` as the model string — the dot is part of the id. This page's URL spells it with a hyphen because URLs cannot carry the dot, but the request body needs the real id.

Is GLM-5.2 free or paid?

Paid. Usage is billed pay-as-you-go in USD from your api.reapi.ai token balance, based on input, output, and any applicable cache-token rates. See the live per-1M-token rates in the pricing table on this page; failed requests are not charged.

Does GLM-5.2 require a monthly subscription?

No. There is no monthly subscription. Add funds to your api.reapi.ai token balance and pay in USD only for the input, output, and any applicable cache tokens each request uses. The live per-1M-token rates are in the pricing table on this page.

How do I get a GLM-5.2 API key?

Sign up at api.reapi.ai and create a key under API Keys, then pass it as a Bearer token in the Authorization header. Chat keys and their token balance are separate from the reapi.ai image and video gateway.

GLM-5.2 — 1M Lossless Context, Open-Weight Coding Flagship

GLM-5.2 is Z.AI's flagship foundation model, built for the era of long-horizon tasks. GLM-5.2 reads a 1M-token context window that Z.AI describes as solid and lossless, writes up to 128K tokens back, thinks by default but lets you switch thinking off, and exposes a reasoning-effort control the rest of the GLM family does not have. On reAPI, GLM-5.2 bills about a third below Z.AI's published per-token rate on both input and output.

Highlights: Thinking mode · reasoning effort dial · JSON output
Best for: Bilingual zh / en tasks, tool agents
Input: Chat messages (OpenAI-compatible)
Output: Text, streaming optional

GLM-5.2modelglm-5.2

GLM-5.2 playground

Chat models run on the api.reapi.ai gateway, which has its own console and its own key. Open the console to run GLM-5.2 through the OpenAI-compatible Chat Completions surface.

Open console Token rates

Pricing

Credit-based — 1 credit = $0.001 USD. Pay only for completed generations.

First test cost

$0.00393.9 credits

One 1K-token prompt with a 1K-token answer at the cheapest tier.

Testing budget guide

Add credits

$10

≈ 2564 tests

$50

≈ 12820 tests

$100

≈ 25641 tests

Input: $0.9; Price · 1M tokens
Output: $3.0; Price · 1M tokens

What is the GLM-5.2 API?

GLM-5.2 is Z.AI's flagship foundation model, built for the era of long-horizon tasks and served on reAPI through an OpenAI-compatible /v1/chat/completions endpoint. It reads a 1M-token context window that Z.AI describes as solid and lossless, writes up to 128K tokens back, thinks by default but lets you switch thinking off, and is the only model in the GLM family that takes a reasoning-effort control. Billing is per 1M tokens, with input and output priced separately.

What you can build with this model

Real-world workflows and production use cases you can build and ship with this model.

GLM-5.2 producing an architecture map from an entire repository

Hand it the whole project, not a folder of excerpts

This is the case Z.AI leads with, and the one the context window exists for. GLM-5.2 keeps module boundaries, architectural constraints, API contracts, directory structure and earlier decisions in view for the length of a task, which is what stops the late stages of a long job from feeling like they were done by someone who never read the beginning. Z.AI's framing is worth quoting: the model does not merely read more context, it carries the engineering judgements it formed earlier into later execution. The way to test that on your own codebase is to ask GLM-5.2 for a technical audit — architecture map, module responsibilities, API contracts, data flows, call chains, technical debt — before you ask it to change anything.

Read the API docs

GLM-5.2 working a staged refactor across many files

Long-horizon refactoring that closes the task

Z.AI describes GLM-5.2 as more stable on cross-file, multi-step, long-chain work: it breaks the goal down, identifies dependencies and risks, then implements, verifies and closes in stages. The named fits are module decoupling, API migration, directory restructuring, SDK adaptation and cross-language refactoring — the jobs that fail not because any single edit is hard but because step forty contradicts step three. Bound the task, state what must not change, and ask GLM-5.2 for the execution plan, impact scope, risk boundaries and verification method before it starts editing.

GLM-5.2 running build, lint and tests and reporting the results

Engineering standards that actually hold under a long run

The failure mode of an autonomous coding agent is not bad code, it is scope creep: an extra dependency, a quietly changed API contract, a skipped test, a commit you did not ask for. Z.AI puts consistency on exactly these constraints — code style, architectural boundaries, dependency rules, build process, testing requirements, commit boundaries — among GLM-5.2's improvements, and specifically under long context and many rounds. Feed GLM-5.2 your real lint rules, build commands, test requirements and prohibited actions, then give it a real task and check the report against them.

Why reAPI for GLM-5.2

About a third below the published rate

Z.AI publishes a per-million-token rate for GLM-5.2 input and output. On reAPI, GLM-5.2 bills roughly a third below both — no subscription underneath it, no minimum commitment, pay-as-you-go in USD against your gateway balance. The exact numbers are in the pricing table on this page.

OpenAI-compatible drop-in

GLM-5.2 is served from a standard `/v1/chat/completions` endpoint. If your code already speaks that shape, adopting GLM-5.2 is a base URL, a key and a model string — the same SDKs work once the base URL points at the gateway. No SDK swap, no vendor-specific client.

One key across GLM, Claude, GPT and Gemini

The same api.reapi.ai key that calls GLM-5.2 calls the Claude, GPT and Gemini families too. Sending the cheap bulk of an agent's turns to GLM-5.2 and escalating only the hard ones to a closed frontier model is a model string, not a second integration, a second invoice and a second set of credentials.

GLM-5.2 vs Kimi K3

These two are the open-weight answer to the same question: a million-token context aimed at long-horizon coding, with the weights published rather than held. GLM-5.2 is the cheaper of the pair on reAPI by a wide margin, and the more forgiving to integrate — thinking is optional, sampling parameters still do something, and prior reasoning is dropped from the history unless you ask to keep it. Kimi K3 answers with modality and a stricter contract. Pick on shape, then on price.

Capability

GLM-5.2 on reAPI

Kimi K3

Weights and positioning

Open — Z.AI reports it as the strongest open-source model on coding and long-horizon benchmarks

Open — the first of the three-trillion-parameter class

Context and output limits

1M-token context described as lossless; 128K maximum output

1M-token context; output defaults lower but can be raised to the whole window

Thinking

On by default and can be disabled; when on, the model decides per request whether to think

Always on and cannot be disabled

Effort control

Accepts seven values that collapse to three real behaviours, defaulting to the deepest

Three genuine rungs, also defaulting to the deepest

Sampling parameters

Temperature and top_p both tunable, with temperature capped at 1.0

Both fixed — sending them returns an error

Input modalities

Text only — vision is a separate model in the lineup

Text, images and video

Comparison reflects behavior documented by each vendor at the time of writing. Rates are described as relationships rather than amounts; the live numbers for GLM-5.2 are in the pricing table above.

Ship GLM-5.2 in three steps

step 01
Create an account and key on api.reapi.ai
Sign up at api.reapi.ai, open the console and generate an API key. That one key reaches GLM-5.2 and every other model on the gateway, billed from a single balance.
Open
step 02
Point your client at the gateway
Set your base URL to api.reapi.ai and your key to the one you just created. Any client that already speaks OpenAI Chat Completions works unchanged — no SDK swap is needed to reach GLM-5.2.
Open
step 03
Send `glm-5.2` as the model
Set the model field to `glm-5.2` — with the dot — and post your messages. Turn on streaming for long outputs, and remember that GLM-5.2 thinks by default at the deepest effort setting, so pick an effort level before you measure cost.
Open

docs/api/glm-5-2

API reference

Drop-in code and the full parameter table.

View full reference

curl https://api.reapi.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "group": "default",
    "messages": [
      { "role": "user", "content": "Hello" }
    ],
    "reasoning_effort": "max",
    "stream": true
  }'

Parameter validation failed

Rejected synchronously with the failing field named. Check enum values, ranges, and URL fields — nothing is charged.

Authentication or balance issue

401 means a missing or invalid Bearer key; 402 means the reserve exceeds your balance. Manage keys and credits in the console.

Content or material rejected

Reference material and outputs pass automated moderation. Rejected tasks fail with a clear error and a full refund.

Task failed or timed out

A task that reaches a failed state is never charged — the reserve refunds automatically. Keep the task id and retry when ready.

Frequently asked questions

Common questions about this model.

GLM-5.2 is billed pay-as-you-go in USD per token against your api.reapi.ai balance, with separate input and output rates and no subscription. Both land about a third below Z.AI's published per-token rate. Current numbers are in the pricing table on this page.

The field defaults to the deepest setting, and this is the detail worth knowing: seven values are accepted for compatibility with other providers, but they collapse. None and minimal make GLM-5.2 skip thinking, low and medium both behave as high, and the level above high behaves as the maximum. So there are three real behaviours, not seven, and picking a middling value does not buy you a middling cost.

Both are open-weight flagships with a million-token context aimed at long-horizon coding, and GLM-5.2 costs a fraction of the other on reAPI. GLM-5.2 is the more forgiving of the two: thinking can be switched off, sampling parameters still work, and earlier reasoning is cleaned out of the history for you. The trade-off is modality — GLM-5.2 is text-only, while Kimi K3 also takes images and video. The versus table on this page has the row-by-row detail.