[태그:] AI
-
Reflection AI’s Beam: A 501B Open-Weight Model for Coding
Reflection AI unveils Beam, a 501B parameter open-weight Mixture-of-Experts model trained with massive-scale reinforcement learning for coding and agentic tasks.
-
ChatGPT Signs Fake New Yorker Cartoons With Real Artists’ Names
ChatGPT is generating New Yorker-style cartoons and signing them with real cartoonists’ pen names, raising new AI attribution and copyright concerns.
-
RemoveMacAI Strips Apple Intelligence Models From macOS 27
RemoveMacAI is an open-source tool that disables Apple Intelligence on macOS 27, deletes its downloaded models, and reclaims disk space.
-
Strata Runs the 125B Qwen3.8-Flash-Next Model on a Gaming PC
Strata is a free, open-source tool that runs the 125-billion-parameter Qwen3.8-Flash-Next AI model on consumer GPUs like the RTX 4090, no server required.
-
Kolibri: Aleph Alpha’s Sovereign Open-Weight German AI Model
Aleph Alpha released Kolibri, a 78B open-weight Mixture-of-Experts model with 3B active parameters, built for sovereign, regulated AI deployments.
-
AWS and Google Cloud Add Hard Budget Caps to Stop Runaway Bills
AWS and Google Cloud now offer hard spending limits that pause usage, addressing fears of runaway bills from coding agents and cloud APIs.
-
Federal Court Blocks Utah’s Anti-VPN Age Verification Law
A federal judge blocked Utah’s SB 73, ruling the anti-VPN age verification law demands technical impossibility and burdens users worldwide.
-
Cloudflare Launches Clef, an Open-Weight Decision Model
Cloudflare released Clef and Clef-flash, open-source decision models on Workers AI, plus an RL fine-tuning platform for agentic workflows.
-
Earendil Ships Pi 1.0, a Minimal Hardened Agent Harness
Earendil released Pi 1.0, a minimal, hardened agent harness, alongside Pi Durable for long-running agentic apps beyond the terminal.
-
Gemini 4 Argon: Google’s New Frontier AI Model Explained
Google unveils Gemini 4 Argon, a frontier AI model for coding, enterprise work, and cybersecurity defense, launching first through a limited cyber defender program.
