GLM-5.3-Flash is Zhipu AI's cost-efficient, natively multimodal model in the GLM-5 family — a 320-billion-parameter Mixture-of-Experts model with 18 billion active parameters, released under the MIT license. It's the model that spent a week topping usage charts anonymously under the stealth codename "Ox Alpha," now confirmed and generally available.
| Developer | Zhipu AI |
|---|---|
| Model ID | z-ai/glm-5.3-flash |
| Release Date | August 26, 2026 |
| Context Window | 1,048,576 tokens (~1M) |
| Max Output Tokens | 131,072 |
| Modalities | Text, image, video (input) → Text (output) |
| Capabilities | Reasoning, streaming, structured (JSON schema) output, tool/function calling |
Natively multimodal
Unlike the text-only GLM-5.3, Flash accepts text, image, and video input, making it a drop-in choice for agentic workloads that need to reason over more than plain text.
MoE efficiency
Only 18B of the model's 320B parameters activate per token, which keeps inference fast and cheap without sacrificing the capability of a much larger model.
Open weights, MIT license
Weights are available on Hugging Face under a permissive MIT license, so the model can also be self-hosted.
1M-token context window
At just over 1,000,000 tokens, GLM-5.3-Flash can hold large documents, long conversation histories, or sizeable codebases in a single request.
High-volume agentic workloads
Its MoE efficiency and low price make it a practical default for agents that need to run at scale without the cost of a full flagship model.
Multimodal reasoning tasks
Native image and video input support means it can be used directly for tasks a text-only model can't handle.
Long-document and long-conversation reasoning
The 1M-token window processes long reports, transcripts, or multi-file context in one pass.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
GLM-5.3-Flash is documented as suited to multimodal assistants and agents, including visual question answering, multi-image reasoning, document and screenshot understanding, coding and tool-using agents, and long-context document intelligence.
It is also documented for coding, reasoning, agentic workflows, long-context text processing, and image or video understanding.
Another documented use is high-volume coding and visual coding/front-end work, including self-checking rendered outputs and iterating on interfaces.
It is further described as fitting browser and GUI workflows, including computer-use tasks and other visual software workflows.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "z-ai/glm-5.3-flash", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.AIMLAPI_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
"model": "z-ai/glm-5.3-flash",
"messages": [
{
"role": "user",
"content": "Hello!"
}
]
}),
});
console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"z-ai/glm-5.3-flash","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
MIT-licensed open weights also available on Hugging Face.
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 41.9 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 71.5 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Terminal-Bench | 84.3 | Autonomous shell/terminal task completion | Source | September 21, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
GLM-5.3-Flash This page | Reasoning + agents | |||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
GLM-5.3-Flash has a 1,048,576 tokens context window and can return up to 131,072 tokens.
GLM-5.3-Flash takes image, text, video as input and returns text.
Use z-ai/glm-5.3-flash as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
GLM-5.3-Flash became available on August 26, 2026.
GLM-5.3-Flash is priced at input $0.195 / 1M tokens, output $0.325 / 1M tokens, cached input $0.039 / 1M tokens.
Yes, GLM-5.3-Flash can stream responses as they are generated.
Yes, GLM-5.3-Flash accepts image input alongside text.
GLM-5.3-Flash was built by Zhipu AI.
