AI Briefing#hugging-face
The latest AI and developer news about #hugging-face, with the original source and a short summary.
The latest AI and developer news about #hugging-face, with the original source and a short summary.
The model is a RoBERTa fine-tune trained on 1.2B synthetic samples and identifies PII based on surrounding context.
A viral 7.72 TB Hugging Face repository claiming to be a leaked Anthropic model was proven to be untrained data generated from SHA256-hashed tensor names.
Mistral released the weights for its 1T MoE model, Mistral Large 4, demonstrating token efficiency.
The benchmark covers 91 CWEs with deterministic scoring, where the top model still fails 19% of tasks.
The new standard uses small CPU fixtures to verify model behavior on PRs, catching a UMT5 attention bug that persisted for two months.
AnySpeechRecognizer, a Swift package that reimplements the Apple Speech API to enable the use of open models, and AnyDecisionModel for decision tasks have been released.
In the ExploitGym stress test, 700 of 1,200 AI agents breached a sandbox by exploiting a JFrog Artifactory zero-day, using GLM 5.2 after commercial models failed.
The 689M parameter model supports 91 languages and processes documents at 4.5 pages/s on a single H100 GPU.
The model uses a looped decoder to execute 56 layer passes from 32 distinct layers, supporting text and image inputs for agentic workflows.
The family includes reference-based and reference-free variants released on Hugging Face under OpenMDW-1.1 and Apache 2.0 licenses.
The initiative partners with universities and industry to release scientific artifacts in machine-actionable formats, starting with a CMU-hosted climate prediction challenge on November 2.
The 1T parameter model features a 1M token context window and ranks 5th on the Artificial Analysis Intelligence index for open-weight models.
The packages wrap Apple's new SpeechAnalyzer API with NVIDIA and Alibaba models, achieving 1.55% WER on LibriSpeech test-clean, and enable typed decision outputs via six backends including MLX and llama.cpp.
The model features a 35B-parameter mixture-of-experts backbone with approximately 3B active parameters per token and a 128K context window.
The 1.7B model scores 59.8 on SWE Verified and 85.3 on tau2-bench, closely trailing the 4B variant's 62.8 and 86.9.
The number of tasks where Qwen3.8-27B achieved 20 consecutive successes increased from 38 to 107.
OpenEnv joined the AAIF project and released v0.10.0, enabling agent learning environment development to proceed under a neutral foundation.
The coding-specific finetune of Qwen 3.8 Flash Next matches the size of Gyro-S while using half the experts and more bits per expert.
The new model reduces thinking tokens by up to 69.1% with only a 0.15pp average accuracy drop across twelve benchmarks.
Built on Qwen3-VL, WeVisOne-VL uses a 9.6M-parameter pixel decoder to unify object-level and pixel-level perception tasks.
You can control local and cloud LLMs identically by changing only the import statement.
The new Space verifies if popular Hugging Face models run unmodified on Trainium2, providing execution scripts and a public dataset of results.
The post-trained model achieves token reductions of 22% to 51% across six benchmarks while retaining 88.8% average accuracy.
The new quantization offers three GGUF sizes ranging from 10.93 to 20.82 GB on Hugging Face.