Cortex Training: Serverless Fine-Tuning for Open-Weight LLMs

| Source: Snowflake Blog

Tags: Snowflake, Cortex Training, fine-tuning, open-weight LLMs, reinforcement learning, serverless GPUs

Snowflake put Cortex Training in public preview, a serverless service for fine-tuning open-weight LLMs with SFT, RL and custom training loops, claiming up to 61% lower cost than competing serverless offerings.

Details

Cortex Training lets teams customize and deploy open-weight LLMs without provisioning GPUs or handling scheduling, checkpointing and distributed execution themselves. It supports familiar training frameworks and low-level serverless APIs for custom training loops, and is built on an open source backend. Snowflake cites benchmarks showing up to 61% lower cost than competing serverless offerings; that figure comes from the vendor. The pitch ties training to the data platform: teams can curate training sets and capture agent traces and logs in Snowflake, creating a feedback loop for post-training models on their own domain data.