Submit Tool

GLM-5.3-FlashX

High-speed serving variant of GLM-5.3-Flash

Paid Language
Visit Product Page →
1000000 context tokens
320B (MoE, 18B active) parameters
MIT license
Sep 2026 released

Z.ai (Zhipu AI) released GLM-5.3-FlashX on September 18, 2026, carrying over GLM-5.3-Flash’s 320-billion-parameter mixture-of-experts design with 18 billion active parameters per token, native multimodality, and hybrid sparse-plus-linear attention. The FlashX variant is tuned for serving speed, targeting throughput up to 200 tokens per second, while keeping the same 1 million token context window.

// similar models in Language

Claude Sonnet 5 The first Sonnet-class Claude model to top Anthropic's flagship writing benchmarks Anthropic · Language · Paid
Claude Sonnet 5.5 Faster Sonnet model for everyday coding and document work Anthropic · Language · Paid
Qwen4-Max Alibaba's flagship proprietary model for its Qwen API and cloud platform Alibaba · Language · Paid
DeepSeek V3 DeepSeek's efficient mixture-of-experts model trained at a fraction of rival costs DeepSeek · Language · Free
Qwen3-235B-A22B Alibaba's flagship open-weight mixture-of-experts model, the cheapest capable frontier model Alibaba · Language · Free
GLM-5 Zhipu's frontier open-weight model, trained entirely on domestic Huawei hardware Zhipu AI · Language · Free