Mixture-of-Experts models all follow one big idea, but they do not route tokens in the same way. This post looks at which routing each MoE model uses by comparing DeepSeek, Qwen, Mixtral, Llama 4, GPT-OSS, Kimi K2, and GLM-4.5 in simple words. You will see how expert count, top-k choice,…
Mixture-of-Experts models all follow one big idea, but they do not route tokens in the same way. This post looks at which routing each MoE model uses by comparing DeepSeek, Qwen, Mixtral, Llama 4, GPT-OSS, Kimi K2, and GLM-4.5 in simple words. You …
Nace AI has open-sourced Drex Max v1, a 27B decision model that can run locally on Apple Silicon, and that is a big deal for developers who want fast private AI on their own machines. The model is built for decisions, not long text generation, so i…
Hugging Face has released Carbon-X, a 14.8B open DNA language model that brings DNA and natural language together for the first time in a practical way. It can read raw DNA, chat in English, and help with tasks like mutation checking, organism clas…
TinyFish makes search free, but that does not mean an AI agent reads the web for free. The real cost comes from two places: the search or fetch request itself, and the much bigger hidden cost of sending all that web content into an AI model as toke…
Sakana AI’s LLM peer review system catches 73% of core-claim errors, and that makes this research stand out in a big way. Instead of asking whether an AI reviewer sounds like a human reviewer, the team asked a much tougher question: can it spot a s…
Alibaba’s Qwen story is one of the fastest growth curves in AI, moving from Tongyi Qianwen, an invite-only chatbot in 2023, to Qwen3.8-2.4T-A95B, an open-weight model with 2.4 trillion parameters in 2026. Along the way, Alibaba kept releasing bigge…
