Narayan Desai

🚀 We are excited to announce the release of MobileLLM-R1 — a series of sub-billion parameter reasoning model. Despite the common belief that reasoning only emerges after training on massive amounts of data, we prove otherwise. With just 4.2T pre-training tokens and a small amount of post-training, MobileLLM-R1 demonstrates state-of-the-art reasoning ability. Here is some highlights: 🎉 100% Open Source: model, training recipe, and data sources. https://lnkd.in/gG-WmYgz 🌟 2x-5x performance boost over other fully open-source models: MobileLLM-R1-950M model scores ~5× higher on MATH vs. Olmo 1.24B and ~2× higher vs. SmolLM2 1.7B—despite being smaller. ⚡ 1/10 of the pre-training token count compared to Qwen: Trained only on 4.2T Tokens total (11.7% compared to Qwen’s 36T tokens), MobileLLM-R1-950M matches or surpasses Qwen3-0.6B on multiple reasoning benchmarks. Collaborating with Ernie, Changsheng, Chia-Jung, Wei, Yuandong, et al. And thanks for the support from Raghuraman, Yangyang, Vikas

539

6 Comments

I've found that most devs have rarely looked at GPU compute dies. There's a tremendous amount of innovation that happens at the silicon level that can help in understanding optimization paths when considering different platforms. This blog, authored by Kyle Aubrey and Nick Stam, takes a deep dive into the Blackwell Ultra GPU architecture including the high-speed interconnects, SMs, and other critical components responsible for driving critical workload performance at scale. Blog link: https://lnkd.in/eUXNFb_m #nvidia #gpus #blackwell #hardware #ai

263

7 Comments

AWS has made real progress on its mission to make the cloud the cleanest and most-energy efficient, while supporting customers on their path to digital transformation. And with our Compute for Climate Change Fellowship, we want to take an additional step by supporting some of the world’s most exciting startups on their journey to develop new climate technologies. This programme will invest up to $4 million in AWS credits to help climate tech startups build solutions that use advanced cloud computing and AI to develop solutions for the climate crisis that is gripping our planet. I’m personally proud that among the 20 global cohort members is AURA-AERO from France, an electric aircraft manufacturer that could reimagine air travel. Through this programme, they will receive guided mentorship from AI and sustainability experts to help them bring their ideas to life and make a real change to the climate. Wishing them all the success on their journey to build a sustainable future with AWS! https://lnkd.in/dp4kPTuM

101

1 Comment

Wild Moose is now consistently hitting >80% accuracy in identifying root causes - within just 3 weeks of deployment. The idea of an AI SRE has been floating around for years now. Investors put their money on it. But customers remained (rightfully) skeptical. Why? Because when something breaks in production, every minute counts. And an AI agent that points you in the wrong direction isn’t just useless - it's harmful. The truth is: building a generative AI agent that understands the intimate dynamics of your system is incredibly hard. That’s why despite all the hype, most tools in this space never made it past demo stage. For us, something shifted earlier this year. Technical bets we made early on - around how to express tribal knowledge, and how to gradually increase the Moose’s autonomy through real-time reasoning - started paying off. Everything started clicking. And with that, something else happened: Our customers became our biggest advocates. It’s hard to describe how energizing that feels as a founder. We’re feeling unstoppable. More to come.

165

17 Comments