Large Language Models

Today's Stories

1. 1

   S&P Global launches dual data retrieval for AI agents

   S&P Global announced Adaptive Retrieval, a new service allowing AI agents and large language models (LLMs) to query multiple S&P Global datasets at once using natural language. It is now available alongside the existing Deterministic Retrieval method through the S&P Global AI Data Portal. As businesses shift from human-driven processes to autonomous AI workflows, AI systems need trusted, properly cited data without extensive engineering work. S&P Global's dual retrieval approach eliminates the complexity of manually finding, validating, and structuring data sources—letting customers focus on building products rather than managing data pipelines.

   The S&P Global AI Data Portal now offers both retrieval methods together. Adaptive Retrieval suits complex, multi-step tasks like research and report generation, while Deterministic Retrieval (built on the Kensho LLM-ready API, available since 2025) handles focused queries such as company research or earnings call analysis. Customers can use one method or both depending on their needs.

2. 2

   S&P Global launches dual data retrieval for AI agents

   S&P Global announced Adaptive Retrieval, a new service letting AI agents and large language models (LLMs) access S&P Global data using natural language queries. This joins the existing Deterministic Retrieval (built on the Kensho LLM-ready API, available since 2025) in a single offering called the S&P Global AI Data Portal, making S&P Global the first to offer both retrieval methods together. As organizations shift from human-driven processes to AI-driven workflows where agents work autonomously, they need data that is properly cited, verifiable, and auditable. S&P Global's dual approach eliminates the complex engineering work historically required to connect high-quality data into AI systems—customers can now access trusted data already structured and ready for AI to use, freeing teams to focus on building products and generating insights rather than preparing data.

   Adaptive Retrieval handles multi-step, complex tasks like in-depth research and report generation across multiple datasets at once, while Deterministic Retrieval suits focused queries (e.g., researching a specific company or analyzing earnings transcripts). Customers can use one method or both depending on their system setup and needs.

3. 3

   Google launches Gemini 3.6 Flash and faster models for AI agents

   Google introduced three new Gemini models—Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber in CodeMender—designed to reduce costs and latency for AI agent applications. Gemini 3.6 Flash reduces output token usage by 17% compared to 3.5 Flash (up to 65% on some benchmarks like DeepSWE by Datacurve) and costs $1.50/1M input tokens and $7.50/1M output tokens. Gemini 3.5 Flash-Lite runs at 350 output tokens per second and costs $0.3/1M input tokens and $2.5/1M output tokens. Developers and businesses running production AI agents need faster, cheaper inference (the step where an AI produces an answer) to scale agentic workflows—systems where AI tools make decisions and take actions autonomously. The new models address this by delivering both efficiency gains and lower per-token costs. Gemini 3.5 Flash Cyber, a specialized cybersecurity model, helps organizations detect and fix code vulnerabilities faster, though it will be available only to governments and trusted partners initially.

   Gemini 3.6 Flash and 3.5 Flash-Lite are available today via the Gemini API, Google AI Studio, Android Studio, Gemini Enterprise, the Gemini app, and Google Search (for 3.5 Flash-Lite). Google is also testing Gemini 3.5 Pro with partners and plans broad availability soon; the company has started pre-training for Gemini 4.

4. 4

   Michaels launches 'Ask Mike' AI shopping assistant on Google Cloud's Gemini

   Michaels unveiled Ask Mike, an AI shopping assistant built on Google Cloud's Gemini Enterprise for Customer Experience, now live on Michaels.com and the iOS and Android apps. Since its May release, the tool has generated nearly 75,000 conversations, with over 60% of interactions focused on product discovery. Ask Mike transforms online shopping from keyword-based search into personalized conversation—customers describe their creative vision and receive curated product recommendations. For Michaels, a retailer centered on creativity and celebration, the assistant moves customers from inspiration to purchase in a single guided dialogue, addressing a core friction point in e-commerce.

   The tool handles a wide range of customer intents—party planning, DIY projects, material sourcing—demonstrating versatility across Michaels' assortment. Google Cloud notes the system went from concept to production in six weeks, suggesting rapid deployment as a competitive model in retail AI.

5. 5

   Anthropic and NEC partner to develop AI capabilities

   Anthropic, the AI company behind Claude, has entered a partnership with NEC to collaborate on AI development and deployment. The collaboration between a leading AI research company and a major Japanese technology corporation signals broader industry momentum toward AI integration across enterprise and infrastructure sectors.

   The specific scope, timeline, and technical details of the partnership remain to be seen in further announcements.