Search results for DeepSeek R1
AI & Enterprise
Power demand surges from a single AI prompt put post-transformer architectures in focus
Surging electricity demand from the spread of AI is exposing the limits of transformer-based large language models and boosting calls for next-generation architectures. With many models still relying on more compute, layers and data, analysts expect data centre spending to rise sharply by 2030. Inference-focused models can consume far more power per prompt. Industry attention is turning to post-transformer designs that activate only relevant parts of a model, aiming to cut inference costs, reduce token use and remain compatible with existing infrastructure.
AI & Enterprise
Chinese AI firms shift from open source to closed models, prioritising profitability
Chinese AI companies such as Alibaba and Zhipu AI are increasingly withholding their latest models from open-source release and focusing on monetisation through cloud platforms, the South China Morning Post reported. Alibaba launched three proprietary models this week that are accessible only via its cloud and chatbot site. The company cited relatively weaker developer interest in its Omni series on Hugging Face. Zhipu AI said customers are shifting from self-hosting to cloud APIs.
Industry
MLPerf inference v6.0 benchmark released as AI inference datacentre chip race intensifies
MLCommons released MLPerf inference v6.0 benchmark results on April 1, with 23 companies submitting 451 results comparing datacentre accelerators from Nvidia, AMD and Intel. The round added new large generative AI models including DeepSeek-R1 and Llama 3.1 405B, and expanded beyond text generation to video and multimodal benchmarks. Nvidia posted top scores with Blackwell-based GB300 and B300 systems, while AMD entered clusters using Instinct MI355X.
