Arctic

Snowflake · April 2024

activeOpen Sourcemixture of expertstext
Parameters480B (17B active)
Context Window4K tokens

Why It Matters

Snowflake's entry into the AI model race, designed specifically for enterprise SQL and coding tasks.

Description

Snowflake's entry into the AI model race, featuring 480 billion total parameters with only 17 billion active per input — one of the most aggressive Mixture-of-Experts sparsity ratios of any model. Designed specifically for enterprise SQL generation and coding tasks rather than general-purpose conversation, reflecting Snowflake's focus on data warehouse and analytics customers.

Key Innovations

MoE
MoEArchitecture where only a fraction of the model's parameters are active for each input, allowing massive scale with lower compute.
Open Weight
Open WeightModel weights are publicly released but training data/code may not be. Enables fine-tuning but not full reproduction.

External Links