NVIDIA Buildmodels.m31a
Models Directory/z-ai/glm-5-3-flash
z-aiUnknownfree endpoint

glm-5-3-flash

z-ai/glm-5-3-flash
NVIDIA Build Card

Multimodal 320B-total / 18B-active MoE with hybrid KDA and sparse MLA attention, native FP8 weights, reasoning and tool calling.

Context Length
1,048,576 tokens
Input Pricing
Not published
Output Pricing
Not published
Recent Latency
No probes

Model Architecture & Limits

Canonical IDz-ai/glm-5-3-flash
Publisher / Originatorz-ai
Model FamilyDeepSeek
ArchitectureTransformer (sparse Mixture-of-Experts, hybrid sparse + linear attention)<br> (GLM-5.3-Flash (`Glm5NextForConditionalGeneration`) — 45 decoder layers)
Context Window Limit1,048,576 tokens
Max Output Tokens4,096 tokens
Endpoint Protocolopenai-chat

Capabilities & Modal Surfaces

Chat Conversation✓ Yes
Tool / Function Calling✓ Yes
Reasoning / Thinking✕ No
Structured Output (JSON)✓ Yes
Vision / Multimodal✓ Yes
Audio Processing✕ No
License & Restrictions
**GOVERNING TERMS:** Use of this trial service is governed by the [NVIDIA API Trial Terms of Service](https://assets.ngc.nvidia.com/products/api-catalog/legal/NVIDIA%20API%20Trial%20Terms%20of%20Service.pdf). ADDITIONAL INFORMATION: Use of the model is governed by the [MIT License](https://huggingfa
Commercial and research use permitted under governing model and API terms.