Open Source AI News - Page 3 of 10
195 articles • Page 3 of 10
Bindu Reddy urges push for open‑source AI to curb monopolies, boost competition
Every tech race ends with a handful of companies owning the track. Bindu Reddy, who runs the model benchmarking platform LiveBench, sees AI heading...
OpenAI releases gpt-oss-120B and gpt-oss-20B under Apache-2.0-style license
For years, OpenAI held its cards close. Now, the company has finally laid a major hand on the table.
Nvidia unveils Vera Rubin platform OpenAI, Anthropic, Meta; adds NemoClaw stack
Jensen Huang is done playing component supplier. Nvidia's new strategy is to own the entire AI pipeline, from the silicon up.
G42 unveils open-source Hindi-English NANDA 87B, built on Llama-3.1 70B MBZUAI
The Hindi-speaking world has waited for an AI that doesn’t just translate, it thinks in the same tongue.
Black Forest Labs releases Flux.2 Apache-2.0, vs Nano Banana Pro, Midjourney
On Thursday, Black Forest Labs dropped Flux.2. The entire family of image generation models is now free for commercial use, released under a...
Large CUDA Tiles Reduce Flash Attention TFLOPS by 18‑43% Across Sequences
Bigger tiles should mean fewer memory accesses, faster attention, right? Not on NVIDIA GPUs.
ACRouter's AI model selector beats Opus-only setups by 2.6x on cost
Enterprises running AI at scale rarely stick to one model. They split traffic across GPT-5.5 for hard reasoning, cheaper open-weight models like Kimi...
Estonian institute benchmarks AI models' vulnerability to Russian propaganda
The Institute of the Estonian Language just graded sixty AI models on a critical new subject: Kremlin propaganda.
z.ai's GLM-5 logs record low hallucination rate, beats Moonshot's Kimi K2.5
Every new AI model claims to beat the last. This one might. Z.ai's open-source GLM-5 just posted a record-low hallucination rate.
SMG releases smg-grpc-proto on PyPI; vLLM integrates via PR #36169
Multimodal AI is slow for a boring reason. Python is in the way. Before any image reaches the brains of a model like Llama 4 Vision, it gets chewed...
Learn to Classify Reviews, Tweets, and Feedback with BERT on Hugging Face
Pretrained models are less a breakthrough and more a fact of work. Nobody trains a language model from zero anymore unless they have a billion...
Arcee launches open‑source Trinity‑Large‑Thinking as Meta steps back from Llama 4
When Meta’s Llama division stumbled in April 2025, Llama 4 hit by quality complaints and whispers of benchmark manipulation, the open-source frontier...
Google open-sources Always On Memory Agent, using SQLite over vector DBs
Google just released an agent with no vector database. None. Its memory is a single SQLite file and the model's own decisions.
Z.ai's Open Source GLM-Image Challenges Google's Nano Banana Pro in Text Rendering
Forget the pretty pictures. The real fight in AI image generation is now over text.
Arcee releases Apache-2.0 Trinity Mini (26B) and Nano Preview (6B) models
Arcee just dropped a gauntlet. Two new models, both under Apache 2.0, a license that signals real openness.
AI 'Slop' Era Lets Agents Create Listings in Minutes, Says Alok Gupta
Scroll through any listing site now. The sameness is palpable, a cheaper feel permeating the digital curb appeal. This is no accident.
Moonshot AI releases FlashKDA with CUTLASS kernels, batching, H2O benchmarks
The race to scale long-context AI has a new contender. Moonshot AI just dropped FlashKDA, a production-grade CUDA kernel that turns their Kimi Delta...
Google Cloud offers managed Slurm to rival CoreWeave and AWS in AI training
Google's cloud division just made a standard move for a company its size: it copied a competitor's idea.
German Commons opens pipeline to free AI datasets from copyright limbo
Building a proper AI for German is harder than it should be. The data you need is either locked up or legally murky.
NVIDIA NeMo powers telco reasoning model for autonomous network workflows
A network operations center runs on speed, precision, and pattern recognition, but every incident is a unique puzzle of fields, close codes, and tool...