Zhipu AI has partnered with Huawei to open-source GLM-Picture, a new-generation picture era mannequin that represents a state-of-the-art (SOTA) milestone in multimodal AI. The mannequin was skilled finish to finish—from knowledge processing to mannequin coaching—utilizing Huawei’s Ascend Atlas 800T A2 {hardware} and the MindSpore AI framework, making it the primary SOTA multimodal mannequin totally skilled on domestically developed Chinese language chips.
In accordance with the businesses, GLM-Picture integrates picture era with giant language mannequin capabilities, enabling unified multimodal outputs. Underneath the API-based utilization mannequin, producing a single picture prices simply 0.1 yuan (roughly USD 0.014). An optimized, sooner model of the mannequin is anticipated to be launched quickly.
The open-sourcing of GLM-Picture highlights ongoing efforts by Chinese language AI builders and {hardware} suppliers to construct a self-sustained AI know-how stack, spanning chips, frameworks, and large-scale fashions.
Supply: Chinastarmarket
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the most recent breakthroughs, get unique updates, and join with a worldwide community of future-focused thinkers.
Unlock tomorrow’s tendencies at this time: learn extra, subscribe to our e-newsletter, and develop into a part of the NextTech neighborhood at NextTech-news.com
