Claude 5: What Changed With Sonnet 5 and Opus 5

Anthropic released Claude Sonnet 5 on June 30, then completed the generation with Claude Opus 5 on July 24. The result is not one model with two speed settings. It is a clearer division of labor between a production workhorse and a model for the most demanding agentic work.

Both are now part of Poly's model lineup, so you can make that tradeoff conversation by conversation without moving your work to another app.

Sonnet 5 Is the Everyday Agent

Sonnet 5 is designed for the work teams repeat: building and reviewing software, operating tools, researching, analyzing documents, and moving through multi-step tasks. Anthropic emphasizes its coding and computer-use improvements, but the more important product change is consistency across an entire workflow.

That makes Sonnet 5 a strong default when you need:

• Fast coding iterations and repository-scale changes
• Tool calls that need to stay on-plan over many steps
• Long document analysis without immediately reaching for the premium tier
• A capable model inside high-volume production workflows

Anthropic launched Sonnet 5 at introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026, before its standard pricing takes effect. Poly displays the current routed price in the model picker so the cost is visible where you make the choice.

Opus 5 Is Built for the Hardest Runs

Opus 5 moves the focus from raw task completion to dependable long-horizon execution. Anthropic describes improvements in coding, knowledge work, computer use, and verification. In practice, that makes Opus the option to try when a task has many dependencies, when an early mistake compounds, or when checking the answer is as important as producing it.

Good Opus 5 use cases include:

• Large migrations that cross many files and systems
• Complex financial, legal, or technical analysis where review matters
• Long-running agents that must preserve intent through many tool calls
• High-value deliverables that justify a slower, more expensive pass

At $5 per million input tokens and $25 per million output tokens, Opus 5 is not the economical default. It is the escalation path.

One-Million-Token Context, Two Different Jobs

Through OpenRouter, both models expose a one-million-token context window and accept text, images, and files. Both also support tools and reasoning controls. That similarity can make the choice look harder than it is.

Start with Sonnet 5. If the task repeatedly loses coherence, needs more rigorous self-checking, or carries a high cost of failure, rerun the critical stage with Opus 5. Poly keeps the conversation in one workspace, making that handoff easy.

The useful pattern is not loyalty to one model. It is using the least expensive model that can reliably finish the job.

Try Claude 5 on Poly

Claude Sonnet 5 and Claude Opus 5 are available to Poly subscribers. Choose Sonnet for the main loop, then bring in Opus when the work demands a deeper final pass.

Start with Claude Sonnet 5 Try Claude Opus 5

Sources: Anthropic's Claude Sonnet 5 announcement and Claude Opus 5 announcement.