Inside Snowflake Intelligence: Five Pillars of Enterprise-Grade Agentic AIExplore the underlying architecture and system-level optimizations behind Snowflake Intelligence, an agentic system built for enterprise.1d ago1d ago
Arctic Long Sequence Training (ALST): Scalable And Efficient Training For Multi-Million Token…Train Llama-8B up to 15M tokens with ALST: open-source, Hugging Face-compatible, no custom code. Up to 469x longer sequences on 4 H100s.Jun 24A response icon1Jun 24A response icon1
Scaling vLLM for Embeddings: 16x Throughput and Cost ReductionHow we improved embedding throughput by 3x in Snowflake Cortex — and pushed to 16x higher compared to vLLM in open source.May 30May 30
Smaller Models, Smarter SQL: Arctic-Text2SQL-R1 Tops BIRD and Wins BroadlyArctic-Text2SQL-R1 by Snowflake AI tops BIRD and wins broadly with a reasoning-first approach, using simple rewards and scalable design.May 30May 30
Arctic Inference with Shift Parallelism: The Fastest Open Source Inference System for Enterprise AIArctic Inference uses Shift Parallelism, SwiftKV, and speculative decoding to power the fastest open-source enterprise AI.May 30A response icon1May 30A response icon1
Fastest Speculative Decoding in vLLM with Arctic Inference and Arctic TrainingA deep dive into how we achieved 4x faster end-to-end task completion for LLM agents and 2.8x faster decoding for open-ended workloads.May 2May 2
Snowflake Arctic Cookbook Series: Arctic’s Approach to DataOn April 24, we released Snowflake Arctic with a key goal in mind — to be truly open. In line with that goal, the Snowflake AI Research…Apr 26, 2024Apr 26, 2024
Snowflake Arctic Cookbook Series: Exploring Mixture of Experts (MoE)On April 24 Snowflake Arctic was released to the world with a key goal in mind — to be truly open. As part of that initiative, the…Apr 24, 2024A response icon1Apr 24, 2024A response icon1
Snowflake Arctic Cookbook Series: Building an Efficient Training System for ArcticOn April 24 Snowflake Arctic was released to the world with a key goal in mind — to be truly open. As part of that initiative, the…Apr 24, 2024A response icon3Apr 24, 2024A response icon3
