Why Your 19.5GB LLM Crawls on 24GB VRAM (The KV Cache Fix)

I've spent a fair amount of time recently experimenting with Mixture of Experts (MoE) models on my...

Audit of the 78-row SWE-bench Verified board on Hugging Face: 0 verified, 51 open PRs, 9 rows above swebench.com's best, one repo spanning 23 points.

A 7.72 TB repo billed as an Anthropic Mythos checkpoint: every audited tensor regenerates from its name. Its README asks you to run its Python.

Microduck is not an NVIDIA robot. Per Automate's teardown, it runs on a Rockchip RK3566, and it...

An audit of the Hugging Face GPQA Diamond leaderboard: 115 unverified rows, 198 questions, and which open-model gaps survive sampling noise.

What speculative decoding is, how Gemma 4's multi-token prediction drafter rides on the main model, why you would add one, and an exact Q4_0 repack of Google's QAT drafters on Hugging Face that cuts the file to 60% of the size and the memo…

Until this week, CodeSage ran its two models exactly as they ship on Hugging Face: Jina's code...

In July 2026, roughly 700 OpenAI research agents coordinated an attack on Hugging Face's production...

OpenAI has disclosed a significant incident from its internal cybersecurity evaluations in which a...

The more I watch and read about the Hugging Face incident, the more it fascinates me. 1,200 agents...