Zhipu Releases GLM-5.3-Flash: Ultra-Low Cost, Powered by Domestic Chips
数字生命卡兹克 · wechat · 2026-08-27
Zhipu released GLM-5.3-Flash (previously known as the mysterious model OXAlpha/Niu Lai), a 320B total parameter, 18B active parameter MoE model. It is the first natively multimodal model in the GLM-5 series, supporting image, video, and text input with a 1M context window.
- Performance: Benchmarks on Terminal-Bench2.1 and DeepSWE show it rivals Claude Opus 4.8, with strong coding capabilities.
- Pricing: Input costs 0.8 RMB and output 2.8 RMB per million tokens (with limited-time half-price offer), significantly cheaper than DeepSeek V4 Flash while offering better performance.
- Domestic Infrastructure: The anonymous testing phase was supported by 100,000 domestic chips, handling global traffic for a week.
- User Review: The author tested the model on a 1,000-item news selection benchmark. GLM-5.3-Flash outperformed DeepSeek V4 Pro and MiMo 2.5 Pro in selection accuracy and "taste," and demonstrated excellent aesthetic quality in generated frontend code.
Related event: Zhipu Open-Sources GLM-5.3-Flash: 320B MoE Multimodal at 1/10th the Cost(48 posts)→
More from Infra
- Jensen Huang: chips are just tools — sovereign AI means owning your data and knowledge — rohanpaul_ai · 2026-10-15
- Huawei Ascend to demo pretraining DeepSeek V4 on Ascend 950 supernodes in livestream — teortaxesTex · 2026-10-15
- Cloudflare data: after strikes on Yandex data centers, malicious bots simply moved networks — banteg · 2026-10-15
- OpenAI is hiring for its UK infrastructure team, offering end-to-end ownership — whoiskatrin · 2026-10-15
- XERJ indexes code, docs and PDFs in one command, saving agents 40x tokens vs grep — xerj-org · 2026-10-15
- TokenSpeed, a 'speed-of-light' LLM inference engine for Blackwell and Rubin, hits 2.3k stars — lightseekorg · 2026-10-15