Overview

GLM 5.3 Flash is the first natively multimodal model in Z.ai's GLM-5 series, featuring 320B total parameters. It introduces a hybrid architecture combining sparse and linear attention, sharply reducing long-context serving costs while preserving precise long-context capabilities. The model also adopts Manifold-Constrained Hyper-Connections (mHC) to further improve scaling efficiency.

Together with Z.ai's 30T-token multimodal pre-training corpus, these changes enable GLM-5.3-Flash to deliver strong intelligence at a low cost.

Parameters

For a more detailed explanation on possible parameters, refer to OpenAI's Create chat completion documentation.

Field	Type	Definition	Default Value
`model`	string	Required. The model key to be used. Use `zai-org/glm-5.3-flash`.	`zai-org/glm-5.3-flash`
`messages`	array	Required. Conversation history. Each object has a `role` (`user` or `assistant`) and `content` (string).	-
`stream`	bool	Routes the request to the streaming path vs non-streaming path.	`false`
`max_completion_tokens`	int	Output token cap.	-
`temperature`	float	Randomness. Greedy sampling at `0` ; identical results are not guaranteed.	`1`
`top_p`	float	Nucleus sampling cutoff.	`1`
`top_k`	int	Limits sampling to the top K tokens.	-
`reasoning_effort`	string	Constrains effort on reasoning for reasoning models. Supported values are `low,` `high`, `max`. Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response.	`max`
`n`	int	How many chat completion choices to generate for each input message. Currently, we only support `1`.	`1`
`presence_penalty`	float	Number between `-2.0` and `2.0`. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model’s likelihood to talk about new topics.	`0`
`frequency_penalty`	float	Number between `-2.0` and `2.0`. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model’s likelihood to repeat the same line verbatim.	`0`
`logprobs`	bool	Whether to return log probabilities of the output tokens or not. If true, returns the log probabilities of each output token returned in the content of message	`false`

Example Requests

from openai import OpenAI

client = OpenAI(
    base_url="https://api-cdn.thehive.ai/api/v3/",
    api_key="<YOUR_SECRET_KEY>",
)
stream = client.chat.completions.create(
    model="zai-org/glm-5.3-flash",
    messages=[
        {"role": "user", 
         "content": "Explain TCP and UDP in two sentences."
        }
    ],
    stream=True,
    reasoning_effort="high",
    extra_headers={"Accept": "text/event-stream"},
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
    if chunk.usage:
        print(f"\n\n[usage] {chunk.usage}")

curl https://api-cdn.thehive.ai/api/v3/chat/completions \
  -H "Authorization: Bearer <YOUR_SECRET_KEY>" \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{
    "model": "zai-org/glm-5.3-flash",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

Note: Virginia (VA1) customers should use the URL: https://api-va1.thehive.ai/api/v3/chat/completions

Multi-Turn Conversations

There is no server-side conversation state — each call is stateless. To continue a conversation, resend the full message history on every request:

messages = [
    {"role": "user", "content": "What's a good name for a pet turtle?"},
    {"role": "assistant", "content": "How about Shelldon?"},
    {"role": "user", "content": "Give me three more."},
]

Response Format (SSE)

The response is a stream of Server-Sent Events, each an OpenAI-style chat.completion.chunk. Content arrives incrementally in choices[0].delta.content:

data: {"id":"...","object":"chat.completion.chunk","model":"<MODEL_KEY>","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

data: {"id":"...","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" there"},"finish_reason":null}]}

The stream ends with a final usage chunk (empty choices, populated usage) followed by data: [DONE] :

data: {"id":"...","object":"chat.completion.chunk","choices":[],"usage":{"prompt_tokens":18,"completion_tokens":42,"total_tokens":60}}

data: [DONE]

Usage is always included - you don't need to request stream_options.include_usage yourself.

Billing

For pricing details, refer to Hive's pricing page under Large Language Models > GLM 5.3 Flash.