Tag archive
Alibaba released its Qwen3.6 line under Apache 2.0, led by a 35-billion-parameter mixture-of-experts model that activates only about 3 billion parameters per token and targets agentic coding.
Investor David Sacks called the AI model layer an 'emerging duopoly' and warned against policies that entrench two firms, opening a public fight over whether open-weight models are a check on concentration or a national-strategy asset.
The UK AI Safety Institute's first public cyber analysis finds leading open-weight models like GLM-5.2 now match closed frontier models from just 4 to 7 months earlier, at a fraction of the cost.
Mozilla's first 'State of Open Source AI' report finds that open-weight models now handle the majority of production tokens on OpenRouter, with the five highest-volume models all open weights.
Moonshot AI's Kimi K3 took the top spot on a frontend-coding leaderboard and helped push AI and semiconductor stocks down for a third straight day, with traders calling it a 'Kimi moment.'
MiniMax M2.7 is the first open-source self-evolving AI agent — 230B MoE, Apache 2.0, 30% self-driven performance gain. Full benchmarks and developer breakdown.
Diffusion text models generate in parallel blocks rather than left to right; Google's open DiffusionGemma and Inception's Mercury 2 are now in a head-to-head over speed.
With a top hosted model pulled overnight, a flood of powerful open models you can run yourself -- and run fast -- is being reframed from hobby to risk management.
Technical Reconstruction of AI Access Dynamics Under Government Restrictions ...
Gemma 4 12B: encoder-free audio and video on 16 GB RAM. Run it via Ollama, HuggingFace Transformers, or llama.cpp GGUF — with fine-tuning no
Kimi K2.7-Code beats Claude Opus 4.8 on MCPMark Verified (81.1% vs 76.4%) at $0.95/M tokens. Open-weight 1T coding model developer guide, June 2026.
I build Atlarix — a desktop coding harness built for the open-weight frontier labs (DeepSeek, Qwen,...