Open-Source AI
Hugging Face, GitHub, Ollama, local models, agent frameworks
The lead
Latest first
Latest in Open-Source AI
GitHub’s Agentic Workflows Now Write Docs Across Many Repos at Once
GitHub has published a detailed walkthrough of how its own engineers use Agentic Workflows to automate documentation that spans many repositories at…
Qwen3: 8 Models, Hybrid Thinking, 119 Languages — Dev Guide
TL;DR — Qwen3 launched April 29, 2025 from Alibaba’s Qwen team. Eight models (six dense: 0.6B–32B; two MoE: 30B-A3B, 235B-A22B), hybrid thinking/non-thinking…
Inside Two Real AI Coding Agents Teams Shipped in 2026
Two real AI coding agents shipped at Hugging Face in 2026: a Slack bot with persistent memory, and a teaching layer that…
Hugging Face Adds One-Command vLLM Servers to HF Jobs
Hugging Face now lets users deploy private, OpenAI-compatible vLLM servers on HF Jobs with a single command, no servers to provision, billed…
GitHub, UNDP Assess Ghana Open Source for Digital Reform
GitHub and UNDP completed an OSPORA readiness assessment of Ghana’s open source governance readiness to support the country’s sweeping digital reform and…
Hugging Face Proposes Cross-Origin Storage API for Transformers.js to Slash Redundant Caching
Hugging Face’s proposed Cross-Origin Storage API for Transformers.js targets 177MB+ duplicate model and Wasm downloads across cross-origin web apps — Hugging…
GitHub details one-year progress on inclusive open source ecosystem pledge
GitHub has published a one-year update on its 2025 pledge to build a more inclusive open source ecosystem, including hackathon results, new…
Hugging Face Data Shows LoRA May Not Be Best Fine-Tuning Option
Hugging Face Data Shows LoRA May Not Be Best Fine-Tuning Option: Hugging Face analysis of 20,834 model cards finds LoRA accounts for…
Hugging Face Benchmark Shows LoRA Isn’t Always the Best PEFT Choice
Hugging Face’s expanded open PEFT benchmark finds LoRA accounts for 98.4% of single-technique Hub model cards, but unoptimized hyperparameters and inconsistent…
Hugging Face PEFT data challenges LoRA’s 98% dominance in fine-tuning
Hugging Face’s PEFT library analysis finds LoRA accounts for 98.4% of single-technique fine-tuning models on its Hub, but notes inconsistent paper benchmarks…