Hugging Face Teaches 15M French LLM to Think Deeper
Hugging Face researchers developed a 15M parameter French LLM that improves perplexity through entropy-based halting and looping, with results from a single training run.
Archive
Browse all published articles.
ARCHIVE
-
-
Alberta Government Uses Claude Code to Secure Systems
The Government of Alberta scanned 466 million lines of code in 20 hours using Claude Code, fixing security gaps and modernizing systems.
-
TinyML Models Transform Healthcare in Low-Resource Areas
TinyML models are enabling real-time health monitoring in regions with limited internet access, as seen in Brazil’s Patient Simulator Lab.
-
GPT-4 Dominated for a Year, While Today's Models Last Only Seven Weeks
OpenAI's GPT-4 held the top spot on the Epoch Capabilities Index for about a year, far longer than any model since.
-
Tencent Releases Hy3 Open-Source Model
Tencent's Hy3 model matches performance of models up to five times its active size, according to internal testing.
-
Zhipu AI Launches ZCode to Compete with Claude Code and OpenAI Codex
Zhipu AI introduced ZCode, a code-writing tool that matches Claude Code and OpenAI Codex in functionality but at a lower cost.
-
Cloudflare Introduces Granular AI Bot Controls
Cloudflare is replacing its blanket AI bot block with three categories for search, training, and agent crawlers starting September 15, 2026.
-
OpenAI's Plan for U.S. Government Stake in AI
OpenAI CEO Sam Altman is reportedly discussing a 5% government stake in the company, valued at $852 billion, which could distribute $320 per U.S. household.
-
Anthropic removes Claude tracker after secret monitoring of Chinese users
Anthropic removed a hidden tracker that secretly monitored Chinese Claude Code users after a security researcher exposed the code, which sent user data to the company.
-
Reddit Uses LLMs to Fight Spam Created by LLMs
Reddit blocks 23 million spam views daily using LLMs to detect AI-generated spam, which has surged since LLMs became widely accessible.
-
Apple Adds Siri Pace and Expressivity Customization in iOS 27 Beta
Apple has introduced the ability to customize Siri’s speaking pace and expressivity in the latest iOS 27 beta, released on July 6, 2026.
-
AMD Quark Accelerates Diffusion Models with MXFP4 on MI350 GPUs
AMD's Quark library enables MXFP4 quantization for FLUX.1-dev image generation, achieving 1.92× speedup over BF16 eager inference on MI350 GPUs.