Announcing Claude Sonnet 5.5 on Snowflake Cortex AI

Snowflake announced the public preview of Anthropic’s Claude Sonnet 5.5 within its Cortex AI platform, accessible initially via Cortex Inference. The model enables teams to perform analysis and application development while retaining data within Snowflake’s security and governance framework. Additional integrations with Snowflake CoWork, Snowflake CoCo, Cortex Agents, and Cortex AI Functions are planned for the near future. The move expands Snowflake’s AI capabilities for enterprise users seeking secure, governed AI workflows.

Anthropic reports that Sonnet 5.5 achieves a 70.6% score on Terminal-Bench 4.0, nearly seven times higher than Sonnet 5’s 10.3%, and performs comparably to Opus 5.5 on GDPval-AA. The model excels in chart recognition and computer use while delivering output over 30% faster than Sonnet 5. It is positioned as a complement to Opus 5.5, ideal for well-scoped tasks requiring speed, such as bug fixes, document creation, and interactive workflows.

Sonnet 5.5 enhances Snowflake CoCo, the AI coding agent, with improved reasoning, code generation, and support for complex, long-running tasks. It enables developers to transition from concept to production quickly while adhering to enterprise governance policies. The model’s capabilities extend to research, structuring applications, and refining outputs through iterative feedback, making it valuable for data engineers and analysts.

Snowflake CoWork, a personal work agent, will also support Sonnet 5.5, allowing knowledge workers to automate routine tasks, answer complex questions with cited research, and execute multi-step workflows securely. The agent leverages Deep Research, proactive automations, and connectors for tools like Gmail, Jira, Slack, and Salesforce, all within Snowflake’s governance perimeter. Support for Sonnet 5.5 in Cortex Agents and Cortex AI Functions is expected to follow in subsequent updates.