GLM 5.3 FlashX
GLM 5.3 FlashX is the high-speed variant of Z.ai's GLM-5.3-Flash, a native multimodal model delivering inference speeds of up to 200 tokens per second. It accepts text, images and video on a 1M-token context window - at $0.37 per million input tokens.
Technical Specifications
Live data fetched from the OpenRouter model registry — always up to date.
What is GLM 5.3 FlashX?
GLM 5.3 FlashX takes the GLM-5.3-Flash line and pushes throughput to as much as 200 tokens per second, built on the same hybrid sparse and linear attention architecture. It is natively multimodal - text, images and video - with a 1M-token context window, and the combination of speed, media input and $0.37 per million input tokens is what sets it apart.
Features & Capabilities
Up to 200 tok/s
Inference fast enough to feel immediate in an interface.
Native Multimodal
Text, images and video handled by the model itself.
1M Context
A million tokens for long documents and media transcripts.
Hybrid Attention
Sparse and linear attention for efficient long context.
GLM 5.3 Capability
Draws on the current GLM generation rather than an older base.
$0.37 per Million
Multimodal speed at a price below the flagship tier.
Who Is GLM 5.3 FlashX For?
Real-Time Interfaces
Output fast enough that users do not wait.
Image and Video Analysis
Media inspection at speed and low cost.
Long Multimodal Context
A large document set plus its media in one call.
Cheap Agents
Many fast steps across an agent workflow.
GLM 5.3 FlashX Pricing
- $0.37 per million input tokens, $1.25 per million output tokens
- Up to 200 tokens per second inference speed
- Run it from your Hoodgen free credits, no credit card required
- Higher allowances on Hoodgen AI paid plans
What Makes GLM 5.3 FlashX Different?
FlashX is the speed tier of the GLM-5.3 family, and the headline is up to 200 tokens per second. On its own that is a latency improvement; combined with native multimodal input and a 1M-token context window it is a different product - media analysis that returns fast enough to sit inside a user interface rather than a batch job.
The hybrid sparse and linear attention architecture is what allows the long context and the throughput together. Attention cost is the usual constraint on both, and mixing sparse and linear attention is the technique that has become standard for getting around it.
At $0.37 per million input tokens it sits well below the frontier tier while offering multimodal input, long context and very high throughput. That makes it a natural default for interactive products and for media pipelines where the frontier models would be prohibitively expensive at volume.
GLM 5.3 FlashX Performance & Pricing
GLM 5.3 FlashX is the multimodal speed tier of the GLM family - faster and cheaper than GLM 5.3 Prime, with media input.
What Can You Do With GLM 5.3 FlashX?
GLM 5.3 FlashX makes sense when responses have to arrive quickly and the input may include images or video. It is the fast, cheap multimodal option in the GLM line.
Best Use Cases
- Latency-sensitive chat and assistants
- Image and video inspection at volume
- Mixed document and media analysis
- Interactive agent workflows
- Replacements for slower multimodal models
GLM 5.3 FlashX Privacy & Data Policy
GLM 5.3 FlashX is developed by Z.ai and served through OpenRouter. Prompts and completions are processed under Z.ai terms.
Since the model accepts images and video, be deliberate about what you upload. Avoid sending confidential recordings or documents you would not share with an external provider.
GLM 5.3 FlashX vs Other Models
Frequently Asked Questions
Multimodal answers at 200 tokens a second
Chat with GLM 5.3 FlashX on Hoodgen - sign in and start in seconds.
Chat with GLM 5.3 FlashX