# What was announced Claude Sonnet 5.5, Anthropic's updated Sonnet model, is available in public preview on Snowflake Cortex AI through Cortex Inference. Calls to the model run inside Snowflake's security and governance perimeter.
# What Sonnet 5.5 is designed to do Sonnet 5.5 targets well-scoped, everyday work where speed and throughput matter. Anthropic positions it as a faster complement to Opus 5.5: Sonnet 5.5 is optimized for focused tasks, while Opus 5.5 remains the choice for complex, open-ended problems that require sustained judgment.
Key performance points reported by Anthropic:
- Terminal-Bench 4.0 score: 70.6% (a large improvement over Sonnet 5).
- Nearly level performance with Opus 5.5 on GDPval-AA, a 44-occupation benchmark for real-world work.
- Substantial gains on chart recognition and computer-use tasks.
- Generation speed >30% faster than Sonnet 5, making it suitable for latency-sensitive or high-throughput workflows.
# Practical use cases inside Snowflake Sonnet 5.5 is recommended for tasks that are well-scoped and multi-step but not open-ended. Typical applications listed by Snowflake and Anthropic include:
- Focused coding tasks: fixing bugs, understanding unfamiliar codebases, making targeted changes, and producing a working first version quickly.
- Content creation: drafting and polishing documents, slide decks, and spreadsheets, especially when following templates.
- Everyday knowledge work across business functions where fast iteration is valuable.
- Visual analysis: reading charts and other visual inputs more accurately than prior Sonnet versions.
- Interactive workflows and agents that need low latency responses.
# How Sonnet 5.5 will be available in Snowflake
- Cortex Inference: Sonnet 5.5 is accessible now through Snowflake's inference endpoint.
- Snowflake CoCo (coming soon): an AI coding agent that will use Sonnet 5.5 for improved reasoning and code generation inside Snowflake-aware workflows and tools.
- Cortex AI Functions (coming soon): integrate Sonnet 5.5 into SQL with functions like AI_COMPLETE so analysts can run natural-language prompts directly in SQL queries.
- Cortex Agents (coming soon): support for long-running agents that sustain complex tasks across many steps.
Example SQL use shown by Snowflake uses AI_COMPLETE('claude-sonnet-5-5', PROMPT(...)) to have Sonnet 5.5 review commentary and compute whether claims are supported by underlying numbers.
# When to choose Sonnet 5.5 versus Opus 5.5 Choose Sonnet 5.5 when your workload is well-scoped, latency-sensitive, or requires high throughput and fast iteration. Choose Opus 5.5 when the task demands complex, open-ended reasoning or sustained judgment. Snowflake and Anthropic present Sonnet 5.5 as a complementary option rather than a replacement.
# What to expect next Snowflake plans broader integration across its AI tooling: CoCo, CoWork, Cortex AI Functions, and Cortex Agents will add Sonnet 5.5 support soon, expanding how developers, data engineers, analysts, and knowledge workers can apply the model within Snowflake's governed environment.
# Bottom line Sonnet 5.5 on Snowflake provides a faster, more capable option for focused coding, content, and routine analytical tasks inside Snowflake's security perimeter. Teams should evaluate Sonnet 5.5 for latency-sensitive and high-throughput workflows and retain Opus 5.5 for more complex work that requires deeper judgment.