Roar AI
DocsRequest access
All models
Zhipu AI · Open weights

GLM-5.3

Z.ai’s open flagship for coding agents and security work.

Input$1.274per 1M tokens
Output$4.004per 1M tokens
BillingOne invoice, in LKRwith every other model
Request early accessSee every model’s rate

About GLM-5.3

GLM-5.3 is Z.ai’s top model, and an unusual kind of upgrade: the base model is the same as GLM-5.2, and every gain came from further training on top of it. The gains are large where Z.ai aimed them — 66.9% on DeepSWE, up from 46.2%, and 84.5% on the CyberGym security benchmark, which Z.ai puts ahead of Claude Fable 5.

It always thinks before answering, at low, high or max effort (max is the default), and handles a million-token context. The weights are open under Z.ai’s own licence, which asks the very largest companies to pass a security review first.

Where it’s strong

Long coding agents that work through a large change.

Security analysis and defensive cyber work.

Command-line automation agents.

Where to be careful

Every call reasons — there is no fast mode — so simple requests are slower and dearer than they need to be.

Z.ai’s headline coding gain is on a private benchmark whose method is not published.

Text only.

Price in rupees

Input$1.274 per 1M tokens
Output$4.004 per 1M tokens
Cached input$0.3276 per 1M tokens

For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $7.83 — a long prompt and a page of answer each. Every rate is published and stays the same from one request to the next. See every model.

Specs

Made byZhipu AI
ReleasedAugust 2026
TypeChat
Context window1M tokens
Longest answer128K tokens
ReadsText
Thinks before answeringAlways on
WeightsOpen (GLM-5.3 License)
SizeAbout 744B total, 40B active
Model idglm-5.3

Using it from Sri Lanka

GLM-5.3 is billed in rupees, on the same monthly invoice as every other model your team uses — no foreign card, no separate Zhipu AI account, and one key for all of it.

Prompts sent to GLM-5.3 are processed outside Sri Lanka. If a workload has to stay in the country, Qwen3.8-27B (Colombo) runs on our own servers in Colombo.

Call it

Use the OpenAI SDK you already have — point it at Roar AI and name the model. Read the docs.

from openai import OpenAI

client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")

reply = client.chat.completions.create(
    model="glm-5.3",
    messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)

Questions

How much does GLM-5.3 cost in Sri Lanka?

Input is $1.274 per 1M tokens; output is $4.004 per 1M tokens. For scale, 1,000 requests of about 3,000 tokens in and 1,000 out cost $7.83. The rate is published and does not change from one request to the next.

Can I pay for GLM-5.3 in Sri Lankan rupees?

Yes. Usage is billed to one monthly invoice in rupees, together with every other model your team uses, and you can also pay by card. There is no separate Zhipu AI account to open and no foreign card needed.

Does GLM-5.3 keep data in Sri Lanka?

No — prompts sent to GLM-5.3 are processed outside Sri Lanka. If your data has to stay in the country, use Qwen3.8-27B (Colombo), which runs on our servers in Colombo.

Is GLM-5.3 open source?

Its weights are open under the GLM-5.3 License licence, so you could run it yourself. Through Roar AI you call it like any other model, with no servers to manage.