SAP is launching its own "tabular" AI model, RPT-1, to compete with general large language models. The model is pre-trained with enterprise knowledge, making it ready for business tasks without extensive training. RPT-1, a Relational Foundation Model, predicts based on relational databases, offering out-of-the-box predictive analytics. It can build business models based on SAP's vast database, and will be available to deploy via SAP's AI Foundation in Q4 2025. Unlike LLMs which learn from text, RPT-1 learns from spreadsheets, providing structured and precise answers. Context engineering enhances the model's performance, by leveraging semantic awareness. The model is built on ConTextTab, a method using semantic signals to enhance training. SAP's model leverages structured business data for tasks with precise answers. SAP’s model differs, as it requires fewer extra details about a business for its predictions.

Google Cloud emphasizes an integrated AI platform for enterprises, combining powerful models with robust operational controls. The company was recognized as a Leader for its Gemini model family in the 2025 IDC MarketScape for GenAI Life-Cycle Foundation Model Software. Gemini's multi-year commitment to building advanced, multimodal AI is a key differentiator. The focus is on optimizing AI workloads for production, prioritizing speed and cost-effectiveness. Gemini 2.5 models offer "thinking budgets" and "thought summaries" for cost control and transparency. Flexibility is enabled through Vertex AI Model Garden, providing access to first-party, open-source, and third-party models. Gemini's native multimodal capabilities, which includes the new Nano Banana, are a core strength. The company invested in coding performance and agentic developer tools, including the Gemini CLI. Vertex AI is presented as a managed platform for scaling AI projects, offering a secure, end-to-end environment. Customization, grounding for accuracy, and security/governance are key components. The IDC MarketScape report provides a comparative assessment of technology vendors in the GenAI landscape.

A turning point for data analysis?

The United Arab Emirates aims to compete with the U.S. and China in artificial intelligence. An Emirati AI lab, the Institute of Foundation Models (IFM), has released K2 Think, an open-source model designed to rival established AI chatbots. Researchers claim K2 Think, with 32 billion parameters, outperforms larger models like OpenAI's ChatGPT and China's DeepSeek in benchmark tests. It reportedly leads all open-source models in mathematical performance and is optimized for math, coding, and scientific research. The model's efficiency promises comparable or better computing power at a lower cost, a strategy similar to DeepSeek's market disruption. IFM emphasizes transparency by open-sourcing not only the models but also their entire development processes. This includes providing researchers with training code, datasets, and model checkpoints. The UAE and other Arab nations are investing in AI to diversify their economies away from fossil fuel dependence. K2 Think represents a significant step in the UAE's ambition to become a major player in the global AI landscape.

Berkeley Lab’s (LBNL) Joint Genome Institute (JGI) has an opening for a Computational Postdoctoral Fellow to contribute to the development,...

Meta has hired two top Apple artificial intelligence executives, Mark Lee and Tom Gunter, to join its Superintelligence Labs team, which focuses on building advanced AI systems. The team is led by Ruoming Pang, who recently joined Meta from Apple with a $200 million pay package. Pang previously led Apple's foundation models team, which developed models used for features like email summaries and Priority Notifications. Lee and Gunter were key members of Pang's team at Apple, with Lee being Pang's first hire and Gunter being a senior member. Meta is aggressively hiring new staff and engineers to keep up with advancements from OpenAI and Google. Apple is trying to retain its engineers by offering raises, but they are still lower than Meta's offers. The departures of Pang, Lee, and Gunter reflect the turmoil at Apple's Foundation Models team. Apple is considering a major strategy change by using external models from OpenAI or Anthropic to power Siri and other Apple Intelligence features. The company is developing versions with both its own models and third-party technology, but has not yet decided which to use as the foundation for Apple Intelligence next year. Apple's own models have been found to be lacking, leading to the consideration of external alternatives.

LLM or custom model: how should you choose the right solution?

Google DeepMind introduced Gemini Robotics On-Device, a vision-language-action (VLA) foundation model designed to run locally on robot hardware. The model features low-latency inference and can be fine-tuned for specific tasks with as few as 50 demonstrations. By Anthony Alford

YouTube video by Discover AI