Model components / Fine-tuning code

Snowflake Cortex Fine-Tuning

Serverless fine-tuning function inside Snowflake Cortex AI for Llama 3 (8B/70B), Llama 3.1 (8B/70B), Mistral 7B, and Mixtral 8x7B. Training and inference both run on Snowflake compute without exporting data to a third-party platform. Picked when training data already lives in Snowflake tables and the value of in-perimeter governance (Unity-style RBAC, audit trails, region pinning across AWS US-West-2/US-East-1/EU-Central-1 and Azure East US 2) outweighs catalog breadth. GA on 2025-02-07; billed per-token in Snowflake credits with no Provisioned Throughput required to serve the tuned model, a differentiator versus Bedrock.

Snowflake Cortex Fine-tuning, fully managed, generally-available PEFT service inside Snowflake. Fine-tunes a fixed set of base models (llama3-8b/70b, llama3.1-8b/70b, mistral-7b, mixtral-8x7b). Token-priced. Confirmed live via Snowflake docs June 2026.

Openness

1 high confidence

1.0

license: Proprietary
service: fully-managed-proprietary(in-Snowflake)
source: closed
method: PEFT(LoRA-class, not specified)
base-models: fixed allowlist
access: Snowflake-account+credits
data: user-data stays in account but engine closed

Proprietary fully-managed SaaS fine-tuning feature; no source, no self-host. Closed.

https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-finetuning recorded 2026-06-04
fully managed proprietary PEFT service, fixed base-model list, credit-based pricing, GA

Adoption

not assessed

No published usage/customer-count figure for the fine-tuning feature specifically; Snowflake has a large data-platform install base but no honest signal for this SKU's fine-tuning adoption. Null.

https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-finetuning recorded 2026-06-04
GA service with usage-history view but no disclosed adoption numbers

Capability

2 medium confidence

2.0

Narrow capability vs the recipe's feature matrix (single PEFT method, fixed older base models, no advanced RL methods). Integrated convenience, not a frontier trainer.

https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-finetuning recorded 2026-06-04
PEFT only, 6 fixed base models, no full-FT/RL methods listed

Unchanged since 2026-07-30 (last edited, not re-checked)