Top News * NVIDIA Jetson AGX Thor now offers up to 7× generative AI performance thanks to software optimizations and supports the latest LLM/VLM models.
* Anthropic’s Claude Haiku 4.5 delivers Sonnet 4‑level accuracy at one‑third the cost and double the speed.
Models & Releases * Anthropic launched Claude Haiku 4.5, a lightweight model costing one-third of Sonnet 4 and running twice as fast, while matching its performance on coding and reasoning benchmarks. * Haiku 4.5 achieved 73% accuracy on SWE‑Bench and 41% on Terminal‑Bench, comparable to Sonnet 4, GPT‑5, and Gemini 2.5.
Hardware & Infrastructure * Apple unveiled the M5 silicon, delivering up to 4× peak GPU AI performance versus M4, featuring a 16‑core Neural Engine and 153 GB/s memory bandwidth. * M5‑powered MacBook Pro and iPad Pro launch with significant AI speed gains. * NVIDIA Jetson AGX Thor software updates provide up to 7× generative AI performance, accelerating LLM/VLM inference. * Nscale signed a deal with Microsoft to deploy ~200,000 Nvidia GB300 GPUs across U.S. and EU data centers, backed by $1.7 B funding. * Arm partnered with Meta to supply its Neoverse platform for scalable AI workloads.
Product Launches * Google’s Veo 3.1 model blends separate images into a single natural‑looking video, adding synchronized audio and richer realism.
New Tools * system_prompts_leaks is an open‑source collection of extracted system prompts from major chatbots, supporting security research. * Klavis provides a MCP integration layer that enables reliable tool usage for AI agents at any scale. * MineContext delivers a proactive, context‑aware AI that blends context‑engineering with ChatGPT Pulse for richer interactions.
Industry & Adoption * AI agents are moving from pilots to enterprise‑wide deployments, accelerating tasks like threat detection, contract review, and crisis response.
The full daily brief: https://aifeed.fyi/briefing
[link] [comments]