Anthropic's Claude Opus 5 now available on Snowflake Cortex AI

Anthropic and Snowflake have announced the public preview availability of Claude Opus 5 on Snowflake Cortex AI, with same-day launch access as part of their partnership. Claude Opus 5 is Anthropic's latest model, characterized as thoughtful and proactive, delivering state-of-the-art performance on coding and knowledge work evaluations. It improves substantially over its predecessor, Opus 4.8, and includes adjustable effort settings that allow users to optimize for either higher intelligence or faster results through token conservation. The model is stronger at verifying its own work and iterating until it succeeds, produces stronger visual outputs, and demonstrates meaningful gains on scientific-research tasks—making it well suited to demanding, long-horizon agentic and analytical work.

Snowflake is integrating Claude Opus 5 across multiple tools within Snowflake Cortex AI. CoCo Desktop, Snowflake's AI coding agent, is now generally available on macOS and Windows and provides developers with integrated access to local files, repositories, terminals, and Snowflake. Unlike general-purpose AI coding tools, CoCo Desktop arrives with the user's Snowflake catalog already loaded and RBAC roles active, eliminating cold starts and governance gaps. With Opus 5, CoCo Desktop enables improved reasoning and code generation while supporting more complex, long-running agentic tasks. For data engineers, analysts, and builders, it transforms natural language into production-ready pipelines, analytics, and AI agents, collapsing weeks of glue work into minutes while remaining governed, interoperable, and enterprise-grade. Users can try Claude Opus 5 in preview soon via the model picker in CoCo Desktop.

Beyond coding, Claude Opus 5 extends Snowflake CoWork, a personal work agent for knowledge workers. CoWork uses Deep Research to decompose complex business questions across structured and unstructured enterprise data, returning fully cited answers in minutes. With proactive automations, it surfaces insights and recommended actions before users ask. Opus 5 enhances CoWork with stronger reasoning, improved performance on long-horizon agentic tasks, and greater consistency across multi-step workflows, enabling every employee to delegate complex investigations and drive real work forward at scale. Cortex AI Functions—which let analysts apply natural language instructions directly in SQL—also benefit from Opus 5, enabling high-performance processing of multimodal enterprise data at lower cost than manually orchestrated AI pipelines. Developers building agentic enterprise applications can use Cortex Inference to leverage Opus 5's adjustable effort levels, balancing reasoning depth and latency per task and integrating with familiar Anthropic tooling through the Anthropic SDK by pointing it to their Snowflake Cortex Inference endpoint. The announcement emphasizes that enterprises want to move agentic AI into production fast, and with Claude Opus 5 available across multiple entry points—from CoCo CLI to Snowflake CoWork, Cortex AI Functions, and Cortex Inference—customers can adopt the capability and accelerate their path to high-impact, production-ready agentic AI at scale.

Snowflake and Anthropic have positioned Claude Opus 5 as a model purpose-built for enterprise agentic work—complex, multi-step tasks that run for extended periods and demand both reasoning depth and consistency. The launch targets a specific pain point: enterprises want to deploy AI agents into production quickly while maintaining security, governance, and data control. By embedding Opus 5 directly in Snowflake's ecosystem, the two companies eliminate friction points that typically slow AI adoption in regulated or data-sensitive organizations. CoCo Desktop, now generally available, exemplifies this approach: it arrives with the user's Snowflake catalog and RBAC roles already loaded, removing the need to re-explain data models or re-authenticate. Similarly, Cortex AI Functions and Cortex Inference let developers build and scale AI pipelines without leaving Snowflake's governed perimeter. The model's adjustable effort settings—a feature highlighted in the announcement—suggest that Snowflake and Anthropic expect customers to optimize for either reasoning power or latency depending on the task, recognizing that not every agentic workflow demands maximum intelligence.

No comments yet. Be the first to share your thoughts!