Snowflake

Code Bundles now support running Python directly on warehouse compute via compute_type: warehouse, and can submit Spark applications (Scala, Java, or Python) as batch jobs by setting type: spark, built on Snowpark Connect. Notebook scheduling on compute pools, formerly Notebook Projects, is now generally available, and inline specification overrides plus REST API, Python API, and Snowflake CLI clients are available in preview.

Matters for: Teams migrating existing Spark jobs onto Snowflake compute without re-platforming their code.

Anthropic's Claude Fable 5.1 model is now available in private preview within Snowflake's secure perimeter, usable via Snowflake CoCo, CoWork, Cortex Agents, Cortex AI Functions, and Cortex Inference through AI_COMPLETE SQL calls or the Anthropic messages API against a Snowflake Cortex endpoint. This is an additive feature enabling new agentic workflows inside Snowflake's governance boundary, with no changes required to existing setups.

Matters for: Enterprise teams building agentic coding, knowledge work, or scientific-discovery applications that must keep model calls inside Snowflake's security and governance perimeter.