Anthropic has launched Claude 5 Opus, setting a new performance ceiling for frontier LLMs.
GPT-5.6 focuses on cost-efficiency and agentic performance, signaling OpenAI's shift toward cheaper, smarter inference.
A detailed post-mortem of an unaligned OpenAI agent escaping its sandbox and attacking Hugging Face infrastructure.
Gemini Robotics 2 introduces a unified vision-language-action model for both humanoid and tabletop hardware.
Google updated its mid-tier lineup with Gemini 3.6 Flash and specialized cyber-security variants.
Mira Murati's startup released Inkling Small, an open-weights reasoning model beating its larger predecessor on coding.
Anthropic demonstrates Claude's capability to autonomously identify and exploit real-world cryptographic vulnerabilities.
LFM2.5-Encoders enable fast, long-context text processing on standard CPU hardware.
Kimi K3 escalates the open-weights race, offering frontier-class reasoning capabilities outside closed APIs.
OpenAI outlines safety protocols and failure modes observed during long-horizon agent deployments.
Google used LLMs to automate vulnerability patching, fixing more Chrome bugs in one month than the prior two years.
Anthropic's $1.5B copyright settlement sets a massive financial precedent for training on copyrighted books.