Zhipu Open Sources GLM-5.3 Flash: Efficient 320B Model Ties Claude Opus
量子位 · wechat · 2026-08-27
Zhipu AI has officially open-sourced GLM-5.3 Flash, the model previously known as the mysterious "OxAlpha" that went viral overseas. Featuring 320B total parameters but only 18B active parameters due to a hybrid linear and sparse attention architecture, the model compresses layers to just 45. It scores 57 on the AA leaderboard, tying with Claude Opus 4.8, but at just 1/40th of the cost. GLM-5.3 Flash is the first natively multimodal model in Zhipu's 5 series, capable of long video understanding, visual coding, and 3D scene construction, with all inference traffic supported by domestic chips. This release marks a significant breakthrough for Chinese models in efficient architecture, open-source ecosystems, and domestic computing adaptation.
More from Models
- Correction: GLM-5.3-Flash features a 1M context window — ArtificialAnlys · 2026-08-27
- Gemini 3.5 Transcribe launches with 85+ languages and function calling — osanseviero · 2026-08-27
- Hands-on: Gemini 3.7 Flash impresses in frontend tasks — doodlestein · 2026-08-27
- Benchmarking Qwen3.8 27B Quantizations: 4-bit Holds Up, 1-bit Collapses — pmigdal · 2026-08-27
- GLM-5.3-Flash Review: 10% Cost, Pareto Frontier Performance — ArtificialAnlys · 2026-08-27
- Google announces pricing details for Gemini 3.7 Flash — OfficialLoganK · 2026-08-27