Changelog Update

Claude Sonnet 4.6 Now Available in Agent Builder

🎯 What is it?

Claude Sonnet 4.6, Anthropic's most capable Sonnet model to date, is now available for you to use when building custom agents in Dust. This latest model brings significant improvements across coding, computer use, long-context reasoning, agent planning, knowledge work, and design—all at the same pricing as Sonnet 4.5.

💡 Why is it useful?

Sonnet 4.6 represents a major leap forward in AI capabilities, bringing Opus-level intelligence at a more accessible price point. Early testing shows that users prefer Sonnet 4.6 over the previous Sonnet 4.5 roughly 70% of the time, and even prefer it to Opus 4.5 (from November 2025) 59% of the time. It's particularly strong at:

• Better coding: Improved consistency, instruction following, and fewer false claims of success

• Computer use: Near human-level capability at navigating complex interfaces and multi-step tasks

• Long-horizon planning: 1M token context window with effective reasoning across entire codebases or lengthy documents

• Agentic workflows: Excellent at complex, multi-step orchestration with tools and data sources

⚙ How does it work?

When creating or editing an agent in the builder, you can now select Claude Sonnet 4.6 as your model. Once stability is confirmed today, it will also become the default model for Dust's global agents, ensuring you get the best performance automatically.

✨ Concrete Use Cases

Here's how you could use it:

Advanced coding agents: Build agents that can review entire codebases, refactor complex logic, and generate production-quality code with fewer iterations—particularly effective for frontend development and financial analysis.

Document analysis agents: Create agents that can read and reason across dozens of research papers, lengthy contracts, or complex enterprise documents in a single request, with improved accuracy on charts, PDFs, and tables.

Workflow automation agents: Deploy agents for multi-step business processes like contract routing, conditional template selection, or CRM coordination—the model excels at branched reasoning and following through on complex instructions.

Computer use agents: Build agents that can interact with specialized software systems, navigate spreadsheets, fill out web forms, and coordinate across multiple browser tabs with near-human accuracy.

📈 Benefits for you

• Better quality at the same cost: Opus-level performance at Sonnet pricing ($3/$15 per million tokens)

• More reliable execution: Significantly reduced hallucinations, better instruction following, and more consistent multi-step task completion

• Greater context capacity: 1M token context window means your agents can work with much larger documents and codebases

• Stronger agentic capabilities: Built specifically for autonomous tool use and complex workflows

🚀 How to access it?

1. Head to the agent builder in Dust

2. When creating a new agent or editing an existing one, select Claude Sonnet 4.6 from the model dropdown

3. Start building—the model is ready to use immediately

Note: Sonnet 4.6 will automatically become the default model for all global agents later today, once we confirm stability in production.

:cite[a41]