Model launchesGoogle launches faster, cheaper Nano Banana 2 Lite image generatorGoogle on Tuesday introduced Nano Banana 2 Lite, a lower-latency, lower-cost version of its in-house image and video generator designed for high-volume, rapid workflows.3 min read
Model launchesAnthropic releases Claude Sonnet 5: a more agentic, lower‑cost midsize modelAnthropic has introduced Claude Sonnet 5, a midsize model with improved agentic abilities that can plan, use tools and run more autonomously than previous versions.3 min read
Model launchesTabFM: a foundation model for tabular prediction using in‑context learningTabFM reframes tabular classification and regression as an in‑context learning (ICL) problem, allowing one‑pass predictions on previously unseen tables without per‑dataset training, hyperparameter tuning, or manual feature engineering.4 min read
Model launchesMeituan releases 1.6-trillion-parameter LongCat-2.0 with 1M-token context and China-made ASIC trainingMeituan has published LongCat-2.0, a 1.6-trillion-parameter Mixture-of-Experts model with a native one-million-token context window, under an MIT license and commercially available pricing.6 min read
Model launchesBase44 launches its own LLM to improve efficiency of vibe-codingBase44, the vibe-coding startup acquired by Wix for $80 million a year ago, has begun rolling out its own large language model, Base1, to better tailor latency, cost and accuracy for app creation.4 min read
Model launchesOrnith-1.0: open-weight language models tuned for agentic codingDeepReinforce released Ornith-1.0, an open-weights (MIT) family of language models optimized for agent-style coding workflows.3 min read
Model launchesAsian startups launch alternatives as U.S. export ban sidelines Anthropic modelsTwo Asian companies — Tokyo-based Sakana AI and Chinese cybersecurity firm 360 — have released AI models they say can stand alongside Anthropic’s restricted Mythos and Fable 5, after a U.S.4 min read
Model launchesOpenAI introduces the GPT-5.6 Sol model, which improves cybersecurity capabilitiesOpenAI announced the GPT-5.6 Sol model, which it says is its most powerful tool to date for cybersecurity tasks; it particularly improves the performance-efficiency ratio for long-term vulnerability research and exploitation.1 min read
Model launchesOpenAI introduces the GPT‑5.6 Sol model with reinforced safety protectionsOpenAI announced the GPT‑5.6 Sol model, for which they have reinforced real-time protection against high-risk cyber activity and repeated abuse; the system was strengthened with human red teaming and…1 min read
Model launchesOpenAI launches limited preview of GPT-5.6 Sol, Terra and Luna modelsOpenAI announced a limited preview of the GPT-5.6 family: the high-performance GPT-5.6 Sol, the GPT-5.6 Terra designed for everyday efficiency, and the cost-effective, fast GPT-5.6 Luna.1 min read
Model launchesOpenAI opens limited preview of GPT‑5.6 family (Sol, Terra, Luna) with phased, government‑coordinated rolloutOpenAI has begun a limited preview of its GPT‑5.6 model family—Sol, Terra and Luna—making the models available initially to about 20 vetted organizations after sharing details with the U.S.6 min read
