Select from a wide-range of easily deployed models from a simple user interface
- ‘AI Starter Kits’ – templates, tools and demos showcasing how to get value from our models for real business use cases like semantic search and domain-assistants
- Simple, API-based integration of models into downstream applications
Pre-trained foundation models
The latest, most powerful open-source models
Honed for your industry and use case
Fine tuned on your data – and owned by you
- Llama 2 variants (7B, 70B): state-of-the-art of open source language models enabling customers to adapt, expand, and run the best LLM models available, while retaining ownership of these models
- BLOOM 176B: the most accurate multilingual foundation model in the open source community, enabling customers to solve more problems with a wide variety of languages, whilst also being able to extend the model to support new, low resource languages
- A new embeddings model for vector-based retrieval augmented generation enabling customers to embed their documents into vector embeddings, which can be retrieved during the Q&A process and not result in hallucinations. The LLM then takes the results to analyze, extract or summarize the information
- A world-leading automated speech recognition model to transcribe and analyze voice data
- Breakthrough multi-modal and long sequence length capabilities
Single, enterprise-wide view of data ingestion, model training, retraining and serving, along with access control and security
- Embedded Kubernetes for scheduling, orchestration and isolation of resources within multi-node cluster
- Authentication, authorization, multitenancy and other enterprise readiness features
- Comprehensive MLOps features to manage model lifecycle (e.g. training and fine-tuning)
- Accessible via web UI, CLI, SDK and REST APIs
Automatically optimizes across SW/HW stack, eliminating need for custom kernel and library development
- Out-of-the-box performance and efficiency
- Dataflow Graphs + Spatial Templates
- Breakthrough accuracy, simplified
- Training and inference optimized for very large language models
Our Dataflow hardware is purpose-built for AI, leveraging large memory essential for the largest, most accurate models. Reconfigurable Dataflow Architecture (RDA) delivers lightning fast performance for training and inference.
- Three tier Dataflow memory for performance and capacity with on-chip memory, high bandwidth memory, and high capacity memory
- 5nm TSMC, 102B transistors
- 104.3 km of wire
- 638 TFLOPS (bf16)
One system for all the AI needs across your organization
Train fast and efficiently
We offer the highest performance for the largest (1T) models, delivering the speed and accuracy enterprises need.
Own your model
Use our powerful foundation models, train on your data – and own your model. Fully owned, fully portable, no lock-in.
Faster time to value
Instant deployability means you can be up and running in days, not years.
Lower cost to build and run
Half the cost of building from a kit of parts. SambaNova’s all in one stack saves money as well as time.
Secure, open and explainable
Our models are built on open standards, ensuring full explainability to regulators and customers, and reliability and consistency in results.
Available & reliable
Demands of large models, lagging performance, scarcity, and high cost of legacy compute systems have restricted access to scaled AI. SambaNova delivers dedicated hardware – always available, reliable, and predictable.
Ready to learn more?
Learn how SambaNova can advance your AI initiatives to help you achieve your impossible.
Stay on Top of AI: trends, information and company news delivered to you.