NVIDIA to Acquire Hugging Face for $12.93 Billion, Delangue Says
Hugging Face CEO Clement Delangue announced the company intends to join forces with NVIDIA in a $12.93 billion acquisition, marking a major inflection point for…
OpenRouter Launches Video Benchmarks for Side-by-Side Model Comparison
OpenRouter introduces video benchmarks that let users compare video models side by side across a variety of challenges. Early highlights: Seedance 2.5 excels at…
When the AI Context Window Turns Into a Horror Meme
A cryptic tweet from OpenCodeJr turns the AI context window into a surreal creepypasta, warning that 'He's getting more long' and urging people to unplug it. It…
MiniMax-M3 Powers HUMAIN-M3, a Frontier Arabic AI Model
MiniMax announces HUMAIN-M3, an Arabic-specialized model built on MiniMax-M3 and further trained on more than 1 trillion Arabic tokens, now available in researc…
NVIDIA CEO Jensen Huang Signals a Future Home for Hugging Face
NVIDIA CEO Jensen Huang announced that NVIDIA will be a great home for Hugging Face, citing the importance of open models for safety, innovation, and sovereignt…
E-Commerce Bench: A New Benchmark for Long-Horizon Autonomous Business Operations
E-Commerce Bench is a new benchmark that tasks AI agents with running online stores for 365 days with ¥100,000, covering sourcing, negotiation, pricing, promoti…
Requesty Adds a Compliance Page for AI Traffic: Enforce ZDR, Regions, and Export Audit Evidence
Requesty's new Compliance page lets teams enforce zero data retention and region restrictions, compare policy vs observed LLM traffic, and export audit-ready PD…
Qwen Steps Into Autonomous Driving with Qwen-Drive-1.0-4B
Alibaba's Qwen team unveils Qwen-Drive-1.0-4B, an Apache 2.0 vision-language foundation model for autonomous driving, covering 3D perception, driving VQA, and m…
Open RL Write-up Released: Environment, Hand-Rated Pool, and All Paintings
A new write-up shares an open RL environment, a hand-rated evaluation pool, three training runs, and every painting generated, with code, models, and dataset av…
Tencent's Sherry Quantization Shrinks Hy4 Preview from 1.5TB to 214GB with Minimal Accuracy Loss
Tencent's Hy4 preview can be quantized from 1.5TB to 214GB using the Sherry method at 1.25-bit weights, with only minor benchmark drops. Quantized GGUF and orig…