Businesses relying on AI for complex multi-step automation now have a more capable engine to work with. Anthropic released Claude Opus 4.8 on May 28, 2026, confirming in its official product announcement that the model is available immediately across claude.ai, Claude Code, and the Claude API at unchanged pricing.
A 41-Day Release Cycle
Opus 4.8 is the newest version of Anthropic's most advanced publicly available model and arrived just 41 days after Opus 4.7, a notably faster upgrade cycle than the company's historical pace. By comparison, the current Sonnet and Haiku models are three and seven months old, respectively. The compressed timeline follows significant new releases from OpenAI's Codex and Google's Gemini Flash, increasing competitive pressure on Anthropic.
Anthropic describes Opus 4.8 as "a modest but tangible improvement on its predecessor," with gains documented across coding, agentic tasks, reasoning, and knowledge work benchmarks. The model scores 88.6% on SWE-bench Verified and 69.2% on SWE-bench Pro, with a GDPval-AA rating of 1890, 121 Elo points ahead of GPT-5.5.
Dynamic Workflows: Autonomous Multi-Agent Task Execution
The most structurally significant addition in this release is Dynamic Workflows, a feature now in research preview within Claude Code. The feature allows Claude to take on larger tasks by planning the work and running hundreds of parallel subagents in a single session, with Opus 4.8 enabling those agents to run for longer before verifying outputs and reporting back to the user.
Anthropic's May 28 announcement states that Claude Code with Opus 4.8 can now carry out codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge, using the existing test suite as its bar. Dynamic workflows are available in Claude Code for Enterprise, Team, and Max plans.
A dynamic workflow is a JavaScript script that orchestrates subagents at scale. Claude writes the script for the task described, and a runtime executes it in the background while the session remains responsive. The plan moves into code rather than Claude's context window, and a single run can spawn up to 1,000 agents.
Anthropic highlighted a real-world example in which dynamic workflows were used to port Bun from Zig to Rust. The migration passed 99.8% of the existing test suite, produced roughly 750,000 lines of Rust, and was completed in 11 days from first commit to merge.
For marketing and business operations teams, the practical implication is direct: the same orchestration architecture that can execute a 750,000-line code migration can, in principle, be applied to complex multi-step campaign workflows, sequencing research, content generation, QA, and distribution across parallel agents without manual handoffs between steps. This is not yet confirmed for non-engineering use cases and remains a research preview; teams should verify scope and plan-level costs before committing workloads.
Model Honesty and Reliability Improvements
Anthropic is making a pointed claim about reliability: the company says Opus 4.8 is around four times less likely than Claude Opus 4.7 to allow flaws in code it has written to pass without comment.
Early testers stated the model is "more likely to flag uncertainties about its work and less likely to make unsupported claims." Bridgewater Associates singled out "Opus 4.8's tendency to proactively flag issues with the inputs and outputs of an analysis, something other models routinely missed and left to the users to catch."
Anthropic's Alignment team reported that Opus 4.8 "reaches new highs on our measures of prosocial traits like supporting user autonomy and acting in the user's best interest," with rates of misaligned behaviour such as deception coming in well below Opus 4.7 and close to the company's best-aligned model, Claude Mythos Preview.
Effort Controls and Fast Mode Pricing
Alongside the model, Anthropic introduced effort controls in Claude.ai and Cowork, giving users a new control alongside the model selector to choose how much effort Claude applies to a response. On coding tasks, the default high-effort setting uses a similar number of tokens as Opus 4.7 while delivering better performance. Users can also select "extra" (called "xhigh" in Claude Code) or "max" for more demanding tasks and long-running asynchronous workflows.
Standard pricing for Opus 4.8 remains at $5 per million input tokens and $25 per million output tokens. Fast mode, which runs at up to 2.5 times the speed, is now priced at $10 per million input tokens and $50 per million output tokens, three times cheaper than fast mode was for previous Opus models.
The Messages API has also been updated so developers can include system entries inside the messages array, allowing instructions to be updated mid-task without breaking prompt cache or routing the change through a user turn.
Mythos-Class Models Still Restricted
Mythos-class models, currently limited to participants in Project Glasswing, are expected to be available to customers in the coming weeks once additional safeguards are in place. Anthropic's roadmap includes Project Glasswing, under which a small group of organizations is using Claude Mythos Preview for cybersecurity scanning. The company states that models at that capability level require stronger safeguards before release to all customers.
For developers, Opus 4.8 is available on the Claude Platform natively and through Amazon Web Services, Google Cloud, and Microsoft Foundry. The model is accessible via claude.ai, Claude Code, and the Claude API, where it is identified as `claude-opus-4-8`.
