Commenters discussed Xiaomi's MiMo v2.6 release, with several praising the team's transparency - rao-v highlighted the realtime training dashboard and detailed tech report - and others calling out specific strengths like affordability (lwansbrough), good latency (esafak), and real-world scientific demos (vatsachak). Technical details and sizes were noted by stymaar and spwa4 (Flash ~309B total/15B activated; Pro ~1.02T total/42B activated) and practical deployment reports came from volf_ (DGX Spark recipe) and syntaxing/pphysch who raised memory and UI design questions. Many commenters (ddxv, MisterMunchkin, joshheitzman) said the models are impressive for open-weight, low-cost options and compared favorably in certain tasks to Western offerings like Claude or Opus, while others (MisterMunchkin, algoth1) pointed out training-data contamination and similarities to DeepSeek.
Opinion divides sharply over benchmarks, trust, and geopolitics. user43928, varispeed and mokre expressed distrust in reported benchmark rankings and fears of bench‑manipulation or nerfs, while algoth1 and verdverm argued Xiaomi didn't cheat and that Chinese labs are rapidly closing the gap with cost-effective techniques (alfalfasprout). Some celebrated increased competition and lower costs (SyneRyder, tacomagick), whereas swingandamiss refused to support Chinese models on geopolitical grounds. UI/UX quirks like the uppercase motif were noted by nemothekid and ddxv, and others debated terminology and framing (jwpapi on “Pareto/Frontier”), leaving the conversation split between enthusiasm for a cheaper capable model and skepticism about metrics and strategic consequences.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
A recent report indicates that nearly half of young people prefer trusting AI over humans for fact-checking. This shift highlights growing confidence in artificial intelligence for verifying information. (news.sky.com)
TypeSafe AI introduced Jev, a new category of decision models that output floating point numbers instead of text, enabling fast and cost-effective classification and scoring tasks. Jev processes unstructured data and answers yes/no, choice, or score questions, making it suitable for applications like spam detection and search ranking. (simonwillison.net)
MiMo-v2.6-Pro is a new AI model focused on balancing intelligence, performance, and cost. It aims to provide advanced capabilities at a competitive price point. (artificialanalysis.ai)
Multiple models of Claude, including Mythos, Fable 5.1, and Opus 5, are experiencing elevated error rates. The incident affects various Claude services and is currently under investigation. (status.claude.com)
Transformers are neural network architectures that use self-attention to process sequences and capture long-range dependencies, revolutionizing AI since 2017. They power models like GPT-2, LLaMA, and Gemini across various domains, with key components including embeddings, attention mechanisms, and MLP layers. (poloclub.github.io)
The Advisory Group on Mathematics and Artificial Intelligence was established at the Institute for Advanced Study to advise AI companies on their interactions with mathematical research. The group aims to promote responsible sharing of mathematical results and serve as a communication channel between mathematicians and the AI industry. (terrytao.wordpress.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.