Kyma APIKyma API
Zhipu AI

Zhipu AI

GLM 5.2

Zhipu's newest frontier-open flagship on Kyma — the model the glm-flagship alias now resolves to — a 744B-parameter mixture-of-experts model with a 1M-token context window, built for long-horizon agentic coding.

Modalities

Text → Text

Below release

fp4 · fp8

Weights

Published

zai-org/GLM-5.2

Where it sits in the catalogue Kyma measures

Every number against every text model Kyma prices per token — a stated rule, not a chosen line-up.

Price · input#39/71

$1.101

output $3.76 · cached $0.29835

median $1.013 · best $0.0494

Price · output#39/70

$3.76

per 1M tokens generated

median $3.531 · best $0.218

Availability · 30d#1/70

100%

580 observations since 2026-06-19

median 96.5% · best 100.0%

Throughput#12/66

73 tok/s

probe median, one fixed prompt

median 36.3 tok/s · best 377.8 tok/s

Response time#32/66

2.67 s

probe median, to a complete answer

median 2.73 s · best 0.84 s

Context

1M

max output 131,072

published by Zhipu AI, not measured here

Tick above each rail is this model, below it the other 70. Dashed rule is the field median, solid is its best. Better is left; the axis stops at the 90th percentile, so a few models sit past its right edge.

Usage

How much this model is actually called here.

Rank

#33

of 100 active models

Tokens served

11.3M

all-time

Platform share

0.3%

of all tokens

Tokens · last 15 daysSep 22 → Oct 6

Two clocks, and why they disagree

Kyma measures this model twice. Both are real; they answer different questions.

Probe · every 6h · 30 days

2.67s to answer

One fixed prompt, on a schedule, to every model. Comparable, because the model is the only thing that changes.

Observations
580
Answered by a substitute
0

Real traffic · last 7 days

3.63s to answer

Your requests, at the lengths clients actually send. Not comparable between models, but it is what running this one feels like.

Requests
99
Completed
100%
p95
5.1 s

The gap is prompt length, not the model degrading. Use the probe figure to choose between models, the traffic figure to budget for your own.

Pricing

Pay per token. Cached input bills at this model’s own cached rate, listed below.

$1.101 /1M input$3.76 /1M output
Hobby10 req/day · 2K in / 500 out
~$1.22/mo
Production1,000 req/day · 2K in / 500 out
~$122/mo
Scale20,000 req/day · 2K in / 500 out
~$2,449/mo
+ Estimate your workload
1,000
2,000
500
30%

Estimated monthly cost

$108

$3.60 / day on GLM 5.2

Same workload on:

GLM 5.3$126+17%
GLM 5.1$203+87%

Estimates use list pricing with cached input at this model's own cached rate. Actual bills depend on real token counts, and every response includes its exact cost.

When to use GLM 5.2

Updated 2026-06-19

Where this model earns its cost — and where it doesn't.

GLM 5.2 is the flagship of Zhipu's GLM line on Kyma and where the `glm-flagship` alias points. Released in June 2026 under the permissive MIT license, it is a mixture-of-experts model with roughly 744B total parameters and ~40B active per token, positioned squarely at coding and agentic workloads.

On the independent Intelligence Index v4.1 it scores 51 — the highest of any open-weight model, ahead of MiniMax-M3 and Kimi K2.6, and ahead of several leading closed models. Through Kyma every call routes with automatic failover across multiple serving paths, so a degraded provider never surfaces as an error, and you reach it with the same OpenAI-compatible key as every other model.

The headline change over GLM 5.1 is context: the window grows from ~200K to a full 1M tokens, with output up to 131K. That holds an entire codebase plus an agent's working history in a single request. Function calling, structured outputs, and extended reasoning are all supported, and implicit prompt caching bills repeated prefixes at a steep discount — which matters for agents that resend a long system prompt on every step.

Long-horizon coding agents

Its core design target — multi-step SWE tasks where the agent reads, edits, tests, and iterates across a repository over a long session.

Whole-repository context

The 1M-token window lets an agent load a large codebase and its own history at once, instead of paging context in and out.

Complex reasoning

Extended reasoning mode works through debugging, architecture decisions, and multi-constraint planning before committing to an answer.

Tool-calling pipelines

Function calling plus structured outputs keep multi-step agent loops parseable and on schema, step after step.

Not ideal for: Image inputs (it's text-only) or latency-sensitive chat UX — it's a medium-speed, premium-priced model that earns its cost on deep agentic work, not quick interactive replies.

How it compares

Against the peers people actually weigh it against.

SpecGLM 5.2GLM 5.3GLM 5.1
Input /1M$1.101$1.178$1.89
Output /1M$3.76$3.702$5.94
Context1M1M203K
ToolsYesYesYes
ReasoningYesYesYes
Throughput73 tok/s28 tok/s28 tok/s

Quick start

Up and running in under two minutes.

  1. 1

    Create an API key

    Sign up and grab a key from the dashboard. GLM 5.2 needs a top-up — the signup credit covers the free tier.

    Get API key →
  2. 2

    Make your first request

    Drop in your key and send a chat completion — fully OpenAI-compatible.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "messages": [
      {"role": "user", "content": "Explain prompt caching in one paragraph."}
    ]
  }'
  3. 3

    Stream responses

    Add "stream": true to receive tokens as they arrive.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

FAQ

Common questions about this model.

What is the context window of GLM 5.2?

GLM 5.2 has a 1M-token context window — roughly 1471 pages of text in a single request.

How much does the GLM 5.2 API cost?

$1.101 per 1M input tokens and $3.76 per 1M output tokens, with cached input at $0.29835/1M — 73% off the input rate on repeated prompt prefixes. No subscription; you pay only for what you use.

Does GLM 5.2 support function calling?

Yes — GLM 5.2 supports tool/function calling and structured outputs (JSON mode), so it works with agent frameworks out of the box.

Are the weights for GLM 5.2 publicly available?

Yes. Zhipu AI publishes GLM 5.2's weights as zai-org/GLM-5.2, so you can download and run the model yourself (https://huggingface.co/zai-org/GLM-5.2, read 2026-08-14). Kyma serves it because it is convenient and has failover behind it, not because it is the only way to reach it.

Is GLM 5.2 ever served below the precision its creator released it at?

Sometimes. GLM 5.2 was released by Zhipu AI at bf16 (https://huggingface.co/zai-org/GLM-5.2/blob/main/config.json, read 2026-08-14), and at least one route serving it here reports fp4 or fp8 — narrower than that. Weights compressed below the release usually answer close to it, but it is not the identical artefact, and nothing in the response tells you which one answered. Kyma checked the routes on 2026-09-21, so it is your call rather than a silent one.

How do I use GLM 5.2?

Kyma is OpenAI-compatible: point your SDK's base URL at https://kymaapi.com/v1, use your Kyma API key, and set the model to glm-5.2. Signing up is free, but this model needs a top-up: the $0.50 signup credit covers the free tier only.

How is GLM 5.2 different from GLM 5.1?

GLM 5.2 expands the context window from ~200K to 1M tokens, raises max output to 131K, and tops the open-weight Intelligence Index. It's the newer flagship, so the glm-flagship alias now resolves to it; GLM 5.1 stays available by its own model ID.

What is the glm-flagship alias?

Kyma aliases let you write integrations that don't hardcode a model ID. Sending model glm-flagship currently resolves to GLM 5.2, and the X-Kyma-Model header on every response tells you exactly which model ran.

Why run GLM 5.2 through Kyma?

One API key and one OpenAI-compatible endpoint cover this and every other model on the platform. You get automatic failover when a serving path degrades, prompt caching discounts on repeated prefixes — a big deal for agents resending long system prompts — and exact per-request cost in usage.cost.

Needs a top-up — the signup credit covers the free tier.Create account →

More models by Zhipu AI

ModelContextInputOutput
Zhipu AIGLM 5.3 Flash1M$0.1791$0.5972
Zhipu AIGLM 5.31M$1.178$3.702
Zhipu AIGLM 5.1203K$1.89$5.94
Zhipu AIGLM 4.7 Flash203K$0.081$0.54
Zhipu AIGLM 4.5 Air131K$0.1945$1.272