On August 27th, news emerged that Zhipu had recently open-sourced its GLM-5.3-Flash (320B-A18B) model. In a rapid response, Moore Threads harnessed the power of its AI training and inference integrated computing card, the MTT S5000, along with its MUSA software stack, to swiftly adapt and efficiently run the model. This achievement represents a significant stride forward in the utilization of domestic computing capabilities to support ultra-large-scale multimodal models.
Earlier, the model, which bears the codename Ox-Alpha (affectionately dubbed 'Niulai' by the community), underwent anonymous public testing on the OpenCode and OpenRouter platforms. Owing to its remarkable inference speed and superior intelligent performance, it swiftly rose to prominence, becoming the most sought-after model of the week and setting new benchmarks for invocation volume on both platforms.