Model launchesEmbeddingGemma 2: 740M-parameter multimodal on-device embedding model for text, code, images, video and audioEmbeddingGemma 2 is a 740 million‑parameter, commercially permissive (Apache 2.0) multimodal embedding model designed for on‑device use.3 min read
Model launchesPlanetary geospatial foundation models improve public health forecasting using Google Earth AI's PDFMGoogle Earth AI's Population Dynamics Foundation Model (PDFM) compresses privacy-preserving signals—aggregated search trends, human mobility and built-environment density, and environmental determinants—into monthly location embeddings that can be plugged into existing epidemiological models.6 min read
Model launchesMusubi releases PolicyLM-1.7B decision model for sub‑50ms real‑time moderationMusubi has published PolicyLM-1.7B, a 1.7-billion-parameter decision model with open weights designed for real‑time content moderation.2 min read
Model launchesSouth Korea to Invest 4.7 Trillion Won in National Effort to Build Leading AI ModelThe South Korean Ministry of Science and ICT announced a 4.7 trillion won (about $3.5 billion) program to develop a world-class artificial intelligence model, with work expected to start in March 2027.2 min read
Model launchesComparison of Anthropic and OpenAI Code Models: Advantages of Opus 5.5Comparison of Anthropic and OpenAI code models between July and September 2026: in July Anthropic performed better, but the difference was small and it was expensive with subscriptions appearing limited ($200).1 min read
Model launchesGPT‑6.1 Sol and Sonnet 5.5 Lead Cost–Performance Shifts as Local Decision Models and Ecosystem Tools Gain TractionOver Oct 1–2, 2026, OpenAI’s GPT‑6.1 Sol and Anthropic’s Sonnet 5.5 dominated independent leaderboard movement by improving performance at lower costs, while local decision-model support and new developer tooling drove much of the community conversation.6 min read
Model launchesPractical guidance for using the GPT‑6 model family efficientlyThis guide outlines practical recommendations for selecting and operating GPT‑6 models—Astra, GPT‑6.1 Sol, and GPT‑6 Luna—so teams can balance capability, latency, and cost.5 min read
Model launchesMicrosoft AI models available on AI Gateway for speech and transcription tasksIn cooperation with AI Gateway, Microsoft AI's secure, transparent and cost-effective language models are available as of October 1, 2026: the 𝙼𝙰𝙸-𝚅𝚘𝚒𝚌𝚎-𝟸.𝟷 speech model for long-form content and…1 min read
Model launchesOpenAI rolls out global ChatGPT virtual try-on and product Favorites featuresOpenAI has launched two new global shopping features for ChatGPT: a virtual try-on that lets users upload selfies or item images to visualize clothing and accessories, and a Favorites library to save discovered products.2 min read
Model launchesAWS open-sources Strands Decider 2B, a compact decision model inspired by TypeSafe's JevAmazon Web Services published Strands Decider 2B, an open-source, low-latency decision model designed to choose between predefined options and report calibrated confidence.3 min read
Model launchesGoogle unveils Argon as flagship of Gemini 4, shares rise 1.5% after announcementGoogle on Wednesday introduced Argon, the flagship model of its Gemini 4 generation, which the company says is its best-performing model to date and compares it with OpenAI's Astra and Anthropic's Opus.2 min read
Model launchesGoogle adds live, speaking avatars to enterprise customer-service chatbotsGoogle has introduced a Live Avatar feature in its Gemini Live platform, allowing enterprise customer-service systems to display real-time animated avatars that lip-sync and respond during voice or text interactions.3 min read
