Input tokens
103 credits
Per 1M input
About $1.85 at the current GLM 5 credit conversion.
GLM 5.3 FlashX is Z.AI’s accelerated serving tier for the 320B-parameter, 18B-active GLM 5.3 Flash model. Compare upstream specs with glm5.app API limits, then try it in chat.
A faster served version of GLM 5.3 Flash for coding and agent workflows.
Z.AI reports 200 tokens per second for FlashX. Actual response speed varies by request and provider.
GLM 5.3 Flash has 320B total parameters with 18B activated per token.
Z.AI lists a 1M-token context window and a 128K output maximum for its model API. Hosted integrations can apply lower limits.
The upstream model and glm5.app public API have different input and output limits.
Upstream model
GLM 5.3 Flash architecture.
Activated per token in the sparse model.
Maximum listed by Z.AI.
Upstream model API limit; hosted APIs can be lower.
Available through the Z.AI model API; output is text.
FlashX is the faster hosted inference option.
GLM 5 public API
Use this ID with the GLM 5 public API.
This API currently does not forward image or video input.
Current cap for this hosted model.
Check model-specific request limits before sending long prompts.
Use reasoning_effort low, high, or max.
OpenAI-compatible text and tool-call requests.
Upstream specifications come from Z.AI. Hosted behavior and limits come from the GLM 5 API model documentation.
GLM 5 API / 04
The public API uses the same GLM 5 credit balance as chat. These are the current hosted API rates.
Input tokens
103 credits
About $1.85 at the current GLM 5 credit conversion.
Output tokens
348 credits
About $6.25 at the current GLM 5 credit conversion.
Performance / 05
These figures come from provider and hosted API documentation, not from an independent benchmark.
Z.AI reported
200 tokens/s
A provider-reported serving speed; individual requests may be slower or faster.
GLM 5 public API
8,192 tokens
The hosted API cap is lower than the upstream model’s listed 128K output maximum.
See both linked sources for the latest figures. Z.AI model documentation · GLM 5 API documentation
Capabilities / 07
Check which capabilities are available through your chosen API before integrating.
Use the hosted model for responsive text generation.
Review code, explain errors, and draft changes from text prompts.
The GLM 5 public API supports function tools with this model.
The upstream model lists a 1M-token context window; request limits vary by integration.
Z.AI supports image and video input. The glm5.app public API currently forwards text and tool calls only.
Use Chat Completions with streaming and an API key.
Comparison / 06
The two options share the Flash architecture and differ in serving speed and access.
Accelerated serving
Uses the same underlying Flash architecture.
Z.AI reported inference speed.
The accelerated serving option.
No distinct FlashX checkpoint is linked in the Z.AI model documentation.
Status reported in Z.AI’s model documentation.
Open model
Total and active parameters.
Upstream model specification.
Standard Flash model code.
Z.AI describes Flash as an open-source model.
Z.AI states it has three times the GLM 5.3 quota.
This compares Z.AI model offerings. glm5.app API rates are listed above.
Use Cases / 08
Use cases for its faster serving tier and tool-capable text API.
Stream code explanations and suggestions while developers work.
Call tools and use their results in multi-step text tasks.
Review large code sections within the model’s available request limits.
Generate text responses with context from support materials.
Use the upstream Z.AI API when image or video input is required.
Build against the glm5.app text and tool-call API with explicit output limits.
Choose chat or the hosted API, then check the limits for your workload.
Select GLM 5.3 FlashX and use your available glm5.app credits.
Send text and tool-call requests with the glm-5.3-flashx model ID.
Confirm credit rates, input support, and output caps in the API documentation.
Integration specimen
This example uses glm5.app’s text-based Chat Completions API with reasoning enabled and streaming output.
Model API documentationcurl https://glm5.app/api/v1/chat/completions \
-H "Authorization: Bearer $GLM5_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.3-flashx",
"messages": [{
"role": "user",
"content": "Optimize this Python async handler for low-latency streaming completions."
}],
"reasoning_effort": "low",
"max_completion_tokens": 1200,
"stream": true
}'Answers about model specifications, hosted API limits, pricing, and access.
It is Z.AI’s accelerated serving tier for GLM 5.3 Flash, a 320B-parameter model with 18B active parameters.
They share the Flash architecture. Z.AI reports 200 tokens per second for the faster FlashX serving option.
Z.AI lists a 1,048,576-token context window. Hosted integrations may impose additional request limits.
Z.AI lists up to 128K output tokens for its model API. The glm5.app public API currently caps FlashX completions at 8,192 tokens.
The upstream Z.AI model accepts visual input, but the glm5.app public API currently forwards text and tool calls only.
Z.AI describes FlashX as a faster serving option. The published open weights are for GLM 5.3 Flash; check its official model page for local deployment details.
Current rates are 103 credits per 1M input tokens and 348 credits per 1M output tokens, about $1.85 and $6.25 respectively. Check the API docs for changes.
Z.AI currently lists GLM 5.3 Flash on the plan and says FlashX is not yet available there. glm5.app uses a separate credit balance.
No. Z.AI lists text output for this model. Use a dedicated image generation model when you need image output.
Open chat with FlashX selected and check your available credits. For API use, read the model documentation before sending a request.
Start here
Explore fast text generation and tool workflows, then check the hosted API rates before integrating.