SambaNova Unveils the World’s Fastest AI Platform, Revolutionizing Developer Access to Llama Models

In a groundbreaking move for AI development, Palo Alto-based company SambaNova Systems has launched SambaNova Cloud, a platform now recognized as the fastest AI inference service in the world. The platform offers unprecedented speed, providing developers with access to Llama 3.1 models at rates that eclipse current industry standards.

This latest offering is powered by SambaNova’s proprietary SN40L chip, allowing Llama 3.1 70B to run at a blistering 580 tokens per second and Llama 3.1 405B, Meta’s most powerful frontier model, at over 100 tokens per second—all in full precision.

The announcement places SambaNova well ahead of its competitors. Compared to its nearest rival, Cerebras Inference, which recently boasted the world’s fastest AI inference, SambaNova’s platform is 25% faster, while also being twice as fast as Groq. In an industry heavily reliant on GPUs, SambaNova’s purpose-built chips are setting new benchmarks for speed and efficiency, redefining what is possible in AI-driven applications.

Andrew Ng, a renowned AI leader and founder of DeepLearning.AI, commented on the significance of this leap in performance, saying it “opens up exciting capabilities for developers building with LLMs.” Ng, whose work has significantly influenced the development of modern AI systems, emphasized how the platform’s speed makes it ideal for agentic AI workflows, which rely on rapid token generation to deliver real-time, dynamic results.

A Breakthrough for Developers Building Agentic AI

SambaNova Cloud is a developer’s dream, offering unprecedented access to the full power of Meta’s Llama 3.1 models, which are among the most popular open-source large language models today. With no waitlist and free API access, developers can start experimenting immediately with Llama 3.1 70B for high-speed tasks or the massive 405B model for more complex, high-fidelity applications.

Rodrigo Liang, CEO of SambaNova, highlighted the versatility of the platform. “Enterprise customers want versatility—70B at lightning speeds for agentic AI systems, and the highest fidelity 405B model for when they need the best results,” Liang said, emphasizing that SambaNova Cloud is the only platform offering both models at these speeds today.

In addition to speed, SambaNova Cloud’s ability to handle 405B at full precision sets it apart from competitors, many of whom rely on Nvidia GPUs and sacrifice precision to improve performance. SambaNova’s chips not only maintain high precision but also significantly reduce the complexity and cost of deploying such large-scale models.

Setting New Records in AI Inference

Independent benchmarks from Artificial Analysis confirmed SambaNova Cloud’s world-leading performance, with Llama 3.1 405B achieving a record-breaking 132 tokens per second. This makes SambaNova the fastest AI inference platform for developers looking to utilize models requiring rapid, high-quality token generation, particularly in real-time applications or agentic AI use cases.

The new platform has already attracted interest from tech companies aiming to enhance their AI-driven solutions. David Keane, CEO of Bigtincan Solutions, expressed excitement over the potential for improved efficiency in their AI-powered sales enablement tools, citing a projected 300% increase in efficiency thanks to SambaNova’s powerful infrastructure.

A Game-Changer for AI Innovation

SambaNova Cloud is available across three tiers, with a free option providing API access to anyone ready to explore the platform’s capabilities. Developers, whether working on cutting-edge agentic AI systems or fine-tuning models for specific tasks, now have a faster, more powerful toolset at their disposal.

As the race for AI dominance accelerates, SambaNova’s latest innovation is poised to shape the future of AI development, offering developers and enterprises alike the speed, precision, and versatility they need to push the boundaries of what’s possible in AI.

With its world-class technology and support for the most sophisticated open-source models, SambaNova Cloud is not just the fastest platform today—it’s the future of AI inference.