open-source
14 articles tagged with this topic
HF Launches Open-Source Robot: The AI Platform War Behind Nvidia's $13B Deal
Hugging Face launches open-source MicroDuck robot; co-founder Thomas Wolf addresses Nvidia's ~$13B acquisition rumors in a Bloomberg interview.
Local AI audio models balloon to 62 — an arena helps you choose
audio.cpp 0.7 ships 62 model families and 85 variants, plus an Arena comparison interface — selection tools signal maturing open-source voice AI.
Local AI Cuts Basketball Edits From 2 Hours to 30 Minutes
A basketball fan built a free YOLO-based tool cutting 2-3 hours of manual highlight editing to 30 minutes — no cloud AI required.
Harvey Stakes Legal AI's Core Delivery on Kimi K3
Harvey ($11B, $350M ARR) launches Tenet via post-training on Moonshot's Kimi K3, serving 2,400 firms and 200K lawyers—vertical AI inflection signal.
$807, One H100, 2.8B Params: Why Small, Specialized Beats Big, General
$807, one H100: Unbounded Labs trained Bart, a 2.8B pre-1931 English LLM, plus the first vintage benchmark. Does small-and-specialized have a future?
Cloned Myself 5x for Client Inquiries — Open-Source Tool, WeChat Off Weekends
Run multiple AI agents at once to handle clients, reply to emails, draft proposals. Open-source, free, easy to customize — saves that 5K/month hire.
Qwen 4B Reasoning Jumps 17% With Almost No Size Increase
ByteOtter's QLAB method lifts Qwen 3.5 4B reasoning from 46.875 to 54.688 at IQ2_XS—a 16.67% gain with only 0.4% size increase.
SenseTime Open-Sources U1.5-Lite: Train Specialized Experts, Ship One Model
SenseTime open-sources U1.5-Lite image model: per-task experts trained separately, distilled into one. Qwen-Image-Bench climbs 13 points to 60.18. Sin
80K-parameter AI model drops — not a ChatGPT rival, but a signal
Independent dev releases Aurora-80K, an 80K-parameter open-source language model. Far below mainstream LLMs, but it shows how small AI models can get.
开源社区冒出个神秘大模型 — 跑分超 Qwen 顶级版，但没人知道谁训的
A MoE model g9v3-39a5b hit benchmarks this week, outscoring Qwen 27B on some tests. llama.cpp port already live. No company, no launch.
Ornith 1.5 Squeezes 9B Model into 5.6 GB — Local AI Deployment Bar Drops Sharply
Ornith Lab releases Ornith 1.5 open-source models — 9B version quantizes to 5.6 GB and runs on consumer laptops. Community benchmarks show 90%+ capabi
Lophius: A Jupyter Workbench for LLM Research — The Missing Tooling Layer Is Local Models' Real
Open-source LLM research tool Lophius launches from the maker of Heretic, folding model inspection, inference, and tokenizer analysis into Jupyter. Th
OpenMythos Reconstructs Claude: Dismantling Commercial AI's Black Box via Papers
GitHub's OpenMythos reconstructs Claude's architecture from public papers. This "paper archeology" forces commercial AI transparency—methodology over
PanicLock: Disable TouchID on MacBook Lid Close for Safety
Open-source Mac tool forces password-only unlock when you close your lid — critical for journalists and privacy-focused builders.