China's AI sector enters post-training era with DeepSeek and GLM rivalry
teortaxesTex · x · 2026-08-16
Discussion highlights the rise of technical founders in China. DeepSeek V4 Pro and Zhipu's GLM-5.3 are compared, with GLM-5.3 praised for using post-training and RL to boost performance without massive parameter increases. The focus is shifting from scaling parameters to efficient training techniques.
More from Companies & People
- Agent sprawl: managing 100 AI agents may be harder than cloud resources — Ok_Intention1336 · 2026-08-16
- Sam Altman's advice on getting your first 100 users remains solid — tibo_maker · 2026-08-16
- Pixel Ads Criticized for Implying Hardware Relies on Gemini to Compensate — gnukeith · 2026-08-16
- US firms in HK may ban own AI models, forcing VPN use — bytebot · 2026-08-16
- Beihang Prof. He Jing: AI Education & Interdisciplinary Research — 量子位 · 2026-08-16
- DeepSeek's Cui Tianyi: From algorithm textbook author to quant expert — teortaxesTex · 2026-08-16