On demand
Building with the Claude 5.5 Family: Choosing the Right Model and Getting More from Every Token
Claude Sonnet 5.5 joins Opus 5.5 on the Claude Platform as a faster, lower-cost complement.
In this session, Anthropic’s research team walks through how we built the 5.5 models and what we found in testing. You’ll learn which workloads still gain the most from frontier intelligence, and which are ready to run faster and cheaper.
Then our Applied AI team puts it into practice. We’ll show how to build evals from your own tasks, use them to choose the right model and hill-climb on quality, and bring cost per task down with effort levels, prompt caching, and orchestration. Send your questions when you register or in chat during the session.
Relatori in evidenza
Lucas Gonzalez
Research Product Manager @ Anthropic
Ben Lehrburger
Member of Technical Staff, Applied AI @ Anthropic
Cosa imparerai
How we built Opus 5.5 and Sonnet 5.5, what we found in testing, and which tasks to run on each
How to compare models by cost per task, including a larger model at lower effort against a smaller one at higher effort
How to build an eval suite from your own tasks, reuse it as new Claude models ship, and hill-climb on quality
How to lower cost per task with effort levels, prompt caching, and orchestration where Opus 5.5 plans and Sonnet 5.5 executes