Summary
Nate Herk puts Claude Sonnet 5.5 head to head with Opus 5.5 on real workflows to show which model fits which job. Both models run at high effort on identical prompts, covering landing pages, messy data turned into multi-view spreadsheets, interactive dashboards, motion design and project planning.
The video starts with pricing. Sonnet 5.5 is listed at $2 per million input tokens and $10 per million output tokens, roughly half of Opus 5.5 at $4 and $20. Herk says Sonnet 5.5 is an upgrade over Sonnet 5, running about 30 percent faster and costing about 30 percent less for most work. He also reviews Anthropic’s own guidance: Sonnet for well-scoped everyday coding, bug fixes, high-volume development and polished documents, and Opus for long-horizon agentic work, careful judgment and the hardest problems.
In the blind comparisons, Herk reviews the outputs before revealing which model made each one, and tracks runtime, tokens and cost. In one planning task, Opus ran about 20 minutes at a little over $9 while Sonnet ran 14 minutes for just under $5, and Opus produced the clearer 90-day plan. The takeaway is to use Sonnet when a task has a clear spec and a way to check the result, and Opus when the goal is more subjective.
📺 Source: Nate Herk | AI Automation · Published September 29, 2026
🏷️ Format: Comparison
