RPLAYReplay प्राइस

On August 26, MINIMAX-W (00100.HK) announced that, according to the main API service agreement, the company agrees to provide API services to businesses operated by Alibaba Group, with a term from the listing date until December 31, 2028. Under the Alibaba Cloud Computing Service Agreement, the company agrees to purchase several cloud products and services from Alibaba Cloud Computing Co., Ltd., with a term from the listing date until December 31, 2028. On August 26 (after trading hours), the company entered into a supplemental agreement to the main API service agreement and a supplemental agreement to the Alibaba Cloud Computing Service Agreement to revise the existing annual cap for ongoing related transactions under the main API service agreement and the Alibaba Cloud Computing Service Agreement for the years ending December 31, 2026, 2027, and 2028.

On August 26, MINIMAX-W (00100.HK) announced that, under the main API service agreement, the company agrees to provide API services to businesses operated by Alibaba Group, effective from the listing date until December 31, 2028. Under the Alibaba Cloud Computing Service Agreement, the company agrees to purchase several cloud products and services from Alibaba Cloud Computing Limited, effective from the listing date until December 31, 2028. On August 26 (after trading hours), the company entered into a supplemental agreement to the main API service agreement and a supplemental agreement to the Alibaba Cloud Computing Service Agreement to revise the existing annual cap for ongoing related transactions under the main API service agreement and the Alibaba Cloud Computing Service Agreement for the years ending December 31, 2026, 2027, and 2028.

According to Beating AI news, after claiming Ox Alpha during the day, Zhipu officially open-sourced GLM-5.3-Flash in the evening, with model weights now available on Hugging Face under the MIT license. The model has a total of 320 billion parameters but activates only 18 billion at a time. It natively supports text, images, and videos, making it the first native multimodal model in the GLM-5 series. The official statement claims that GLM-5.3-Flash outperforms GLM-5.2 overall, with programming and agent evaluations nearing Claude Opus 4.8, yet it is priced at only one-tenth of GLM-5.2. It also features a new foundational model, introducing a mixed architecture of sparse and linear attention for the first time in the GLM main series, pre-trained on 30 trillion tokens of multimodal data. Previously, community tests on Ox Alpha's DeepSWE small sample reached 80%, but that was based on only 10 questions. After expanding the sample size, the score dropped to about 63%, and testers have proactively corrected the initial claims. This score still places it in the top tier, but it is not as exaggerated as the initial 80%. With the weights open-sourced, developers can directly deploy using frameworks like vLLM, SGLang, and KTransformers without needing to go through the anonymous model's API.