What's New with Claude Opus 4.6

• Agentic coding reaches a new level: Claude Opus 4.6 delivers stronger planning, longer-running agentic behavior, and improved self-review, making it especially effective for working within large, complex codebases.
• Long-context unlocks new workflows: The introduction of a 1M-token context window enables developers to reason over entire projects, documentation sets, and extended tasks without constant context loss.
• Production-ready with DigitalOcean: Accessing Opus 4.6 through Claude Code or DigitalOcean Gradient AI Platform via Serverless Inference allows teams to quickly integrate its capabilities into real applications, dramatically reducing development time while maintaining reliability and control.

Claude Opus 4.6 is a true step forward for the Claude Opus family of Large Language Models. They report that the model is showing serious improvements in general knowledge, coding ability, agentic ability, and multidisciplinary reasoning. Let’s outline some of those new features:

• Expanded long-context capabilities: For the first time in an Opus-class model, Opus 4.6 offers a 1 million token context window (in beta), enabling the model to reason and maintain coherence over much larger codebases, documents, and extended tasks.
• Improved performance and reliability: Building on its predecessor, Opus 4.6 delivers stronger coding performance, more sustained agentic behavior on complex tasks, and better code review, debugging, and planning across extended workflows.
• New agentic and workflow features: The update introduces support for agent teams (parallel task handling), enhanced context compaction for long-running agentic workflows, and adaptive thinking and effort controls for fine-tuning performance and cost.

Together, these improvements on an already fantastic model elevate Opus to stand as one of the best in this generation of LLMs. Now, let’s take a closer look at using Opus.

Getting started with Claude Opus 4.6 with Claude Code is easy! All you need is a Anthropic Claude Pro account, and access is automatically granted by default. Simply follow the instructions on their Getting Started page and authenticate your device using the instructions provided by Claude Code to get started.

Once you have activated Claude Code on your machine, navigate to the directory you want to work in. For this example, we are using the realtime speech translator we released last week for this tutorial. Type claude into the terminal in that directory to get started.

In this example, we asked Claude Opus 4.6 to clean up the application and update its visual theme to better align with DigitalOcean’s branding. We began by having the model propose a set of improvements, then applied those changes step by step rather than all at once. Its suggestions included adding clearer descriptions and centering the web app layout. We found this iterative approach to be more effective than requesting multiple changes simultaneously, as it helped reduce hallucinations.

Above, you can see the original appearance of the web app, and below is the result after applying the requested updates.

As we can see, this is a significant improvement for the UI, and the changes reflect the requests we made. While hundreds of lines were changed, we only had to make three inputs to the model to make these changes. This significantly reduced the time it would have taken to complete the same task on our own.

The other way to access Claude Opus 4.6 is with DigitalOcean Gradient AI Platform via Serverless Inference! Serverless Inference allows you to directly plug the AI into whatever functionality you are using with just a few lines of code. Create a model access key right now by clicking on the link here.

Launching with DO allows you to package all of the features of DigitalOcean’s ecosystem with the power of Claude Opus 4.6 directly. We can use this to enhance our agents with real-world context by creating custom knowledge bases or connecting external data. Additionally, built-in connectors make it easy to ingest files from AWS S3, Dropbox, and more, without complex configuration.

Claude Opus 4.6 represents a meaningful inflection point for agentic AI, especially for developers building complex, real-world systems. With its expanded context window, stronger long-horizon reasoning, and markedly improved reliability in large codebases, Opus 4.6 feels less like a reactive assistant and more like a true collaborative engineer. Paired with tools like Claude Code and platforms like DigitalOcean Gradient, it dramatically compresses the time between idea and implementation: turning multi-hour refactors, reviews, and UI improvements into tightly guided, iterative workflows. As models like Opus 4.6 continue to push the boundaries of what’s possible, DigitalOcean is well-positioned to provide the infrastructure and tooling needed to turn that raw capability into production-ready impact.