Claude Opus 4.8

What Is Claude Opus 4.8?

Claude Opus 4.8 is the latest and most capable AI model from Anthropic, the San Francisco-based AI safety company. Released on May 28, 2026, it builds on its predecessor, Opus 4.7, with meaningful improvements in coding, agentic tasks, reasoning, and overall reliability. Anthropic describes it as "a more effective collaborator" — and early testers seem to agree.

What makes Opus 4.8 stand out isn't just raw performance on benchmarks (though it has that too). It's the model's judgment. In real-world use, Opus 4.8 is noticeably better at catching its own mistakes, flagging uncertainties, pushing back when a plan doesn't make sense, and proactively pointing out issues that other models silently overlook. Anthropic's own evaluations show it's around four times less likely than its predecessor to let flawed code pass without comment.

You can access Claude Opus 4.8 through claude.ai (Anthropic's chat interface), Claude Code (a command-line coding tool), Cowork (a desktop automation tool), and the Anthropic API. It's also available through Amazon Bedrock, Google Vertex AI, and Microsoft Foundry for enterprise deployments.

Key Capabilities

Agentic Coding

Coding is where Opus 4.8 truly shines. It scored 88.6% on SWE-bench Verified and 69.2% on the harder SWE-bench Pro, outperforming both GPT-5.5 and Gemini 3.1 Pro on most coding benchmarks. But the numbers only tell part of the story. In practice, Opus 4.8 is better at understanding large codebases, planning multi-step changes, and executing them cleanly. It asks the right questions before diving in, catches errors in its own output, and handles complex multi-service architectures with more confidence than previous versions.

Reasoning and Analysis

Opus 4.8 delivers richer, more information-dense analysis compared to earlier models. Whether you're working through a research question, analyzing financial data, or breaking down a complex business problem, the model produces outputs with a better signal-to-noise ratio. Testers noted it finishes faster while providing deeper insights — and it has a useful habit of proactively flagging potential issues with both inputs and outputs that users might otherwise miss.

Writing and Knowledge Work

For writing tasks — drafting reports, creating content, summarizing documents, or crafting emails — Opus 4.8 maintains Claude's reputation for clear, natural prose. It handles long documents effectively thanks to a 1-million-token context window on the API (200K on some platforms), meaning you can feed it entire codebases, lengthy contracts, or full research papers and get coherent analysis back.

Honesty and Safety

Anthropic has always emphasized safety, and Opus 4.8 pushes the bar higher. It scores at new highs on measures of prosocial behavior like supporting user autonomy and acting in the user's best interest. It's also less likely to make unsupported claims or engage in deceptive behavior — rates of misaligned behavior are lower than Opus 4.7 and comparable to Claude Mythos Preview, Anthropic's most advanced internal model.

Tool Use and Agentic Workflows

Opus 4.8 uses tools cleanly and follows instructions with exceptional consistency. This matters most for agentic use cases — scenarios where the model operates semi-autonomously, chaining together multiple tool calls, browsing the web, executing code, and managing multi-step workflows without constant human oversight. Early enterprise testers reported that Opus 4.8 is reliable enough to run unattended workloads.

New Features Launched Alongside Opus 4.8

Dynamic Workflows

Available in Claude Code for Enterprise, Team, and Max plans, dynamic workflows let Opus 4.8 tackle very large-scale problems. The model can plan work and spawn hundreds of parallel subagents in a single session, enabling codebase-scale migrations across hundreds of thousands of lines of code. This is a research preview feature, but it represents a significant step toward AI handling genuinely complex engineering projects.

Effort Control

Users on claude.ai can now control how much effort Claude puts into a task. Need a quick answer? Set effort low. Working through a complex problem that deserves deep thinking? Crank it up. This gives you practical control over the speed-versus-depth tradeoff in everyday use.

Fast Mode

Fast mode lets Opus 4.8 produce tokens at roughly 2.5 times its normal speed. Anthropic made this dramatically more affordable — at $10 per million input tokens and $50 per million output tokens, it's three times cheaper than fast mode was for Opus 4.7. This makes it viable for latency-sensitive applications where speed matters.

Who Is Claude Opus 4.8 For?

Opus 4.8 serves a wide range of users. Software developers and engineering teams use it through Claude Code for writing, reviewing, and refactoring code — especially complex, multi-file changes across large projects. Researchers and analysts lean on it for deep reasoning, document analysis, and synthesizing information from large bodies of text. Business professionals use it for writing, strategy, and knowledge work through the claude.ai interface. And developers building AI-powered applications integrate it via the API to power chatbots, agents, research tools, and automated workflows.

If you're someone who needs an AI that can handle serious, complex work — and that you can trust to flag problems rather than hide them — Opus 4.8 is built for you.

How to Use Claude Opus 4.8

Through Claude.ai

The simplest way to use Opus 4.8 is through claude.ai, Anthropic's web and mobile chat interface. Sign up for a free account or log in with an existing one. On the free plan, you'll have access to Claude Sonnet 4.6 with daily usage limits. To access Opus 4.8, you'll need a Pro, Max, or Team subscription. Once subscribed, select the Opus 4.8 model from the model picker and start chatting. You can upload documents, images, and code for analysis, use web search, and organize your work into Projects.

Through Claude Code

Claude Code is a command-line tool that lets developers delegate coding tasks to Claude directly from their terminal. Install it, authenticate with your Anthropic account, and point it at your codebase. Opus 4.8 can read your project files, understand the architecture, make multi-file changes, run tests, and iterate based on results. It's available on Pro, Max, Team, and Enterprise plans.

Through the API

For developers building applications, the Anthropic API provides direct access to Opus 4.8 using the model identifier claude-opus-4-8. The API supports a 1-million-token context window, 128K max output tokens, tool use, adaptive thinking, and all the features that make Opus 4.8 powerful for agentic applications. It's also available through Amazon Bedrock, Google Vertex AI, and Microsoft Foundry.

Pricing

Claude Opus 4.8 is accessible through several pricing tiers depending on how you want to use it.

The Free plan on claude.ai gives you access to Claude Sonnet 4.6 with daily usage limits. Opus 4.8 is not available on the free tier.

Claude Pro costs $20 per month and includes access to Opus 4.8, Sonnet 4.6, and Haiku 4.5, along with roughly 5 times the usage capacity of the free plan, Claude Code access, and features like Cowork and deep research.

Claude Max comes in two tiers: $100 per month for 5 times Pro capacity, and $200 per month for 20 times Pro capacity. These plans are designed for heavy users — developers, researchers, and professionals who work with Claude for hours every day.

Team plans start at $25 per seat per month and include shared workspace features, admin controls, and centralized billing. Premium Team seats at $100 to $150 per seat add Claude Code and higher limits.

For API access, Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens in standard mode. Fast mode costs $10 input and $50 output per million tokens. Batch processing offers a 50% discount, and prompt caching can reduce input costs by up to 90%.

Supported Platforms

Claude Opus 4.8 is available on the web at claude.ai, through iOS and Android mobile apps, a desktop app for macOS and Windows, Claude Code (terminal-based), and Cowork (desktop automation). For enterprise and developer use, it's accessible through the Anthropic API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry.

Pros and Cons

Opus 4.8's greatest strengths are its reliability and judgment. It catches mistakes other models miss, flags uncertainties instead of making things up, and handles complex multi-step tasks with consistency that makes it suitable for unattended workflows. The 1-million-token context window is among the largest available. Fast mode at 3 times cheaper than before makes high-speed use much more affordable. And the deep integration with Claude Code makes it one of the best AI models available for software development.

On the other hand, Opus 4.8 is only available to paying subscribers on claude.ai — free users don't get access. API pricing, while competitive with GPT-5.5, is still at the premium end of the market. And while it beats most competitors on coding and knowledge work benchmarks, GPT-5.5 edges it out on some terminal-based and CLI workflow tasks.

Final Thoughts

Claude Opus 4.8 is a refinement rather than a revolution — but it's a refinement that matters. The improvements in judgment, honesty, and agentic reliability make it a model you can genuinely trust with complex, high-stakes work. For developers, the combination of strong coding performance, dynamic workflows, and affordable fast mode makes it a compelling choice. For professionals and teams, the blend of deep reasoning, long-context understanding, and natural communication makes daily work noticeably smoother. If you're already in the Claude ecosystem, upgrading to Opus 4.8 is an easy decision. If you're evaluating AI assistants for the first time, the Pro plan at $20 per month is a solid entry point to one of the most capable models available today.