SambaNova Systems

Agents are the future—but enterprises hit 3 roadblocks: Security. Speed. Cost. ...We’re fixing all three! Today, we've open-sourced a Deep Research Framework to empower enterprises with: ✅ Your Data, Your Rules: Build agents on your private data, deploy securely on-prem with SambaNova. ✅ 3X Faster Than GPUs: Blazing-fast inference via SambaNova Cloud—iterate faster, act quicker. ✅ Save Millions/Year: Ditch OpenAI’s costs. Use open models like AI at Meta's Llama 3.3 70B for GPT-4o performance at a fraction of the price. How? The framework ships with 3 ready-to-go agents: 1️⃣ General Search Agent (quick answers) 2️⃣ Deep Research Agent (detailed reports) 3️⃣ Financial Analyst (market insights) 🎥 Watch SambaNova smoke GPUs in our demo. 💡 Code your own agent, join the #AIGameOn contest, and win prizes! 👉 Try the FREE demo or clone the repo to start building. The future of enterprise AI is open, fast, and yours. #OpenSourceAI Learn more 👇 https://lnkd.in/gnWxpTxZ

🏆 LET'S GO! 🙌 Thrilled to be named one of Fast Company's Most Innovative Companies for 2025, taking the Number 4️⃣ spot in the Computing Category! 🦾 Our team is driven by a passion for innovation and a desire to disrupt the status quo. We’ve built the best chip for AI Inference that’s power-efficient and a game changer for solutions customers. We’ve built the best chip for AI Inference that’s power-efficient and a game changer for solutions customers. We're honored to be part of this prestigious list — thank you, Fast Company! #FCMostInnovative #AIInnovation #FutureOfWork Adam Bluestein Brendan Vaughan

We love to see it! 🚀 The fact that this entire experiment runs in under 2 minutes with SambaNova Cloud powering the inference is *chef's kiss* 🤌 Excited to see more of what everyone can do with this open-source pipeline 🙌 #AI #OpenSource #AgenticAI

In honor of #WomensHistoryMonth, we want to shine a light on our Chief People Officer, Dana Bennett. 💜 Thank you, Dana, for making sure SambaNovans continue to thrive as we continue to grow and innovate in the AI space! Swipe through to see Dana's journey as CPO at SambaNova. 👇 #WHM2025 #WomenEmpowerment 

DeepScaleR—a compact yet powerful 1.5B parameter model fine-tuned from Deepseek-R1-Distilled-Qwen-1.5B using simple reinforcement learning (RL). Developed by Sijun Tan (UC Berkeley) and team, this model smashes benchmarks with a 43.1% Pass@1 accuracy on AIME2024—a +14.3% leap over its base model! 💡 Why it matters: Despite its small size, DeepScaleR outperforms OpenAI’s o1-preview, proving that smarter training can unlock huge potential in leaner architectures. Efficiency meets excellence! 👏 Shoutout to the team for pushing the boundaries of what’s possible with RL! Dive into the details: https://bit.ly/4kmLiRf

We're making waves across the globe 🦾 "SambaNova ... helped us work with Aramco, one of the energy leaders in the world today, on making sure that we realize, one of the real use cases of AI within Agentic AI, a billion dollars worth of savings." In a guest lecture at Stanford University's CS153, H.E. Abdullah Alswaha, Saudi Arabia's Minister of MCIT, shouts out how SambaNova has been a game-changer for aramco, unlocking new possibilities and major savings 💰 Watch his lecture👇  https://lnkd.in/g42traAP #AI #GenAI #SaudiArabia

DeepSeek R1 671B is ridiculously good, but running at 158 tokens per second is something out of this world. A model *this good* running *this fast* is 3 years ahead of this time. The secret: DeepSeek in this short video isn't running on GPUs. It's running on custom-made chips! This is running on SambaNova and their SN40L chips, which are designed and optimized for running AI workflows. SambaNova is sponsoring my work. We designed GPUs to run games and later realized we could use them for AI. This is very different from creating a chip specifically for AI from the ground up. Just think about this for a second: One single SNL40 chip can simultaneously hold 100+ models (trillions of parameters) in memory! An agentic workflow that uses multiple models simultaneously won't need to swap models from memory even once! By the way, the Llama family of models *screams* on the SambaNova Cloud. Here are some numbers from my latest tests: • Llama 3.3 70B: 367.32 t/s • Llama 3.2 1B: 2381.82 t/s • Llama 3.2 3B: 1335.71 t/s • Llama 3.1 8B: 929.58 t/s • Llama 3.1 70B: 369.16 t/s • Llama 3.1 405B: 103.37 t/s You can access all of these models for free! Create a free account on SambaNova and use Llama 3.1 8B, 70B, and 405B for free using the link below: https://fnf.dev/4broV96 They also have a QuickStart guide with several examples for you to start using their API: https://lnkd.in/eRpt-z6P

Guess who's back? 😎 The team is going to be at the hottest AI Agent dev meetup in Silicon Valley once again! ⚡️ Fast Inference 💻 Open Source AI 🦾 Agentic AI ... we're talking about it all! Join us 👇  https://lu.ma/rm5l38en

POV: You want to build a deep research agent from scratch 🤯 Aurimas Griciūnas is the real MVP for showing us how to build a Deep Research Agentic system with SambaNova in this SwirlAI newsletter 🦾 He outlines step-by-step how he achieves this, from defining the state of the entire system to producing the report. Check it out 👇 https://lnkd.in/gDU69C6u #AI #GenAI #DeepResearch

💡 Ethics in AI Begins with Controlling Your Data 🎙️ Our CEO Rodrigo Liang, recently joined Daniela Braga of Defined.AI and Krystal Hu of Reuters for a panel discussion on AI and ethics at the HumanX Conference in Las Vegas. The conversation started on ethics and the regulatory environment but quickly widened, touching on trust, transparency, open source, and Sovereign AI. 🤖 If you think of our journey with AI as a walk down the yellow brick road from The Wizard of Oz, there’s that moment when it’s time to pull away the curtain to see who’s behind it and what they’re doing. It’s a good metaphor for AI. The more we pull back the curtain, the more we can trust it and ensure it’s being used for the greater good. 🏆 Transparency As a Competitive Edge One key takeaway from the conversation was the importance of transparency in AI systems. Rodrigo talked about SambaNova’s commitment to open-source models because they provide a level of transparency that's essential for building trust. He said that in the coming months, trustworthiness will be a new competitive edge for AI services. 🤔 Of course, this raises questions about the role of regulation in AI development. Daniela pointed out that the current regulatory landscape is fragmented and often unclear, leading to a "wild west" mentality. Rodrigo said he thinks this is changing and that we're seeing a shift toward more responsible and transparent AI development. As more AI players offering similar capabilities emerge, customers will place a premium on trust and transparency. 🦾 Sovereign AI: Taking Control of Your Models But transparency is just the beginning. Rodrigo explained SambaNova’s approach to Sovereign AI, which is our name for the ability of organizations to own and control their own AI models rather than relying on third-party providers. This is particularly important for highly regulated industries like finance and healthcare and some governments, where data privacy and security are paramount. ⚡️ By using open-source models as a foundation and fine-tuning them with their own private data, organizations can create their own AI systems tailored to their needs and compliant with local laws. He said that SambaNova is “meeting the market” where organizations want to use data they've accumulated over decades. 📈 It’s Your Data. Use it! He spelled out SambaNova’s message to those organizations: “Use the data where it is. We’ll bring the AI to you,” including hardware and software and models that they can use just like they would chat with ChatGPT only in a private and secure way. Ultimately, data and all the rules governing its use, including ethics and privacy becomes the essence of differentiation. “It's your data,” he said. “We’re trying to encourage people to leverage it and deploy it in a way they can control, make sure you’re doing the right things with it.” #HumanX2025