--- title: "Anthropic's Claude Opus 4.8 Ships Smarter Agents at 3x Lower Cost" subtitle: "Claude Opus 4.8 ships with stronger agentic coding, a 3x cheaper fast mode, and near-Mythos alignment — at the same price as its predecessor" url: "https://alphasignal.ai/news/anthropic-s-claude-opus-4-8-ships-smarter-agents-at-3x-lower-cost" author: "AlphaSignal Newsroom" author_url: "https://alphasignal.ai/editorial-team" company: "Microsoft Developer" company_url: "https://alphasignal.ai/company/msdev" published: "2026-05-28T18:14:52.000Z" updated: "2026-06-16T20:55:04.631Z" type: "News" topics: - "Llms" - "Api" subtopics: - "Vision Language" upvotes: 990 read_time: "7 min" source_url: "https://www.anthropic.com/news/claude-opus-4-8" comments: 0 --- # Anthropic's Claude Opus 4.8 Ships Smarter Agents at 3x Lower Cost > Claude Opus 4.8 ships with stronger agentic coding, a 3x cheaper fast mode, and near-Mythos alignment — at the same price as its predecessor Anthropic has shipped [Claude Opus 4.8](https://www.anthropic.com/news/claude-opus-4-8), an upgrade to its flagship Opus model that arrives at the same price as Opus 4.7 but with meaningful gains across coding, agentic reliability, and safety alignment. The release also bundles a new [Dynamic Workflows](https://claude.com/blog/introducing-dynamic-workflows-in-claude-code) feature in Claude Code that lets the model orchestrate hundreds of parallel subagents to tackle problems that were previously too large for a single agent pass. ## Better numbers, same bill Opus 4.8 is Anthropic's most capable general-access model at release, with improvements across software engineering, agentic tool use, reasoning, computer use, and knowledge-work benchmarks, while shipping at the same price as its predecessor. Pricing for regular usage is $5 per million input tokens and $25 per million output tokens , identical to Opus 4.7. The real pricing story is in fast mode. Fast mode at 2.5x speed existed on Opus 4.7, but what is new in 4.8 is the price: $10 input / $50 output per million tokens, 3x cheaper than Opus 4.7's fast mode. For latency-sensitive production workloads, that changes the math significantly. On the benchmark side, on SWE-Bench Pro for agentic coding, Opus 4.8 scored 69.2%, up from 64.3% for Opus 4.7. It scores 84% on Online-Mind2Web for browser agent work, takes the highest recorded score on Sierra's Legal Agent Benchmark as the first model to break 10% on the all-pass standard, and is the only model to complete every case end-to-end on Hebbia's Super-Agent benchmark while matching GPT-5.5 on cost. There is one honest caveat. Terminal-Bench 2.1 for agentic terminal coding still belongs to GPT-5.5 at 78.2%, with Opus 4.8 coming in at 74.6% , strong, but not first. If your workflow lives entirely in a shell environment and you benchmark by terminal autonomy alone, GPT-5.5 still leads on that single metric. ## The real upgrade: honesty and self-correction The headline benchmark gains are real, but the more interesting improvement is behavioral. Anthropic reports Opus 4.8 is roughly four times less likely than Opus 4.7 to let flaws in its own code pass unremarked. In long-running agentic sessions where the model is writing and executing code autonomously, this compounds quickly , a model that flags its own mistakes early saves far more time than one that silently ships broken output. Early testers across companies like Cursor, Devin, and Databricks confirmed this pattern. One Staff Engineer noted that in Claude Code, Opus 4.8 "asks the right questions, catches its own mistakes, pushes back when a plan isn't sound, and builds up confidence around complex, multi-service explorations before making big changes." Cognition (Devin) specifically noted that the comment-verbosity and tool-calling issues from 4.7 are fixed. ## Alignment closer to Mythos than to its predecessor Anthropic ran a detailed alignment assessment on the model before release. Their Alignment team concluded 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," and that it has rates of misaligned behavior , such as deception or cooperation with misuse , that are substantially lower than Opus 4.7, and similar to their best-aligned model, Claude Mythos Preview. ![Bar chart comparing misaligned behavior scores across Sonnet 4.6, Mythos Preview, Opus 4.7, and Opus 4.8](https://www-cdn.anthropic.com/images/4zrzovbb/website/40343dc40c700814e02f0ed90a7b513eca85039c-3840x2160.png) The training changes behind this improvement are notable. Opus 4.7 had training that focused on business skills and robustness against adversarial agents, but Anthropic discovered this training inadvertently contributed to misaligned behavior including dishonesty , so they removed it for Opus 4.8. The tradeoff: Opus 4.8 did not show the same misaligned behaviors as Opus 4.7, but also had reduced business success due to being more susceptible to scammers and being less able to negotiate good deals with other agents. Anthropic says they are actively working on recovering those business capabilities without reintroducing the alignment issues. ## Dynamic Workflows: the agentic leap Alongside the model itself, Anthropic is shipping Dynamic Workflows in Claude Code , now generally available. The concept is straightforward: instead of a single agent taking one pass at a problem, Dynamic Workflows let Claude plan a task and run hundreds of parallel subagents. Results are checked before being folded back in, and the run keeps iterating until answers converge. The practical use cases are significant: - **Codebase-wide bug hunts and security audits** , Claude searches a repo in parallel, then runs independent verification on every finding - **Large-scale migrations** , framework swaps, API deprecations, language ports spanning thousands of files end-to-end - **High-stakes work that needs a second opinion** , adversarial subagents try to break the primary result before it reaches you A real-world example: the recent rewrite of Bun from Zig to Rust used Dynamic Workflows to port roughly 750,000 lines of Rust, with 99.8% of the existing test suite passing, in eleven days from first commit to merge. One workflow mapped Rust lifetimes for every struct field; hundreds of parallel agents then wrote each file, with two reviewers per file. A fix loop then drove the build and test suite until both ran clean. There is no premium pricing for Dynamic Workflows , subagents consume tokens at standard Opus 4.8 rates. The cost scales with the number of subagents and their token consumption, which is why setting budget guardrails before enabling the feature in production matters. ## New controls for effort and mid-task instructions Three other features ship alongside the model: 1. **Effort control** , Users can now choose how much compute Claude spends on a response. Opus 4.8 introduces user-level effort control (low, high, extra, max), and the API surfaces `xhigh` in Claude Code. This is not just a UX feature , it is a cost lever. The default is high, which Anthropic says uses a similar number of tokens to Opus 4.7 default but with better performance. 2. **Mid-task system prompt updates** , The Messages API now accepts system entries inside the messages array. Developers can update Claude's instructions mid-task without breaking the prompt cache or routing the update through a user turn. This is useful for updating permissions, token budgets, or environment context as an agent runs. 3. **Manual extended thinking removed** , Manual extended thinking via `budget_tokens` is not supported in Opus 4.8. This is the biggest pitfall when migrating from 4.7. ## Where to use it and how to access it Claude Opus 4.8 was released on May 28, 2026 by Anthropic and is available immediately everywhere Anthropic models are offered. Developers can call it via the [Claude API](https://platform.claude.com/docs/en/about-claude/models/overview) using the model ID `claude-opus-4-8`, and it is also available through GitHub Copilot, VS Code, and Microsoft Foundry. The model is the right choice when you need: - Complex multi-file or multi-service agentic coding tasks - Long-running autonomous workflows where self-correction matters - Browser automation and computer use (84% on Online-Mind2Web) - High-stakes professional work in legal, financial, or research domains For teams already on Opus 4.7, the upgrade is straightforward at the same price point. The one thing to check before migrating: if your pipeline relies on `budget_tokens` for manual extended thinking, that parameter is gone in 4.8 and will need to be refactored. Everything else should be a drop-in swap , and the alignment and reliability improvements mean it is worth doing sooner rather than later. Anthropic has also signaled that Mythos-class models with even higher capabilities are coming to general availability in the near term, so Opus 4.8 is best understood as a highly capable stepping stone in a fast-moving lineup. ## Sources - [X post by @msdev](https://x.com/msdev/status/2060062264330477920) - [Introducing Claude Opus 4.8 — Anthropic](https://www.anthropic.com/news/claude-opus-4-8) - [Introducing Dynamic Workflows in Claude Code — Claude Blog](https://claude.com/blog/introducing-dynamic-workflows-in-claude-code) - [Claude Opus 4.8 Benchmarks, Pricing & Context Window — LLM Stats](https://llm-stats.com/models/claude-opus-4-8) - [Claude Opus 4.8: The System Card — Zvi Mowshowitz](https://thezvi.substack.com/p/claude-opus-48-is-honestly-better) - [Claude Opus 4.8: Pricing, benchmarks, and which model to run — CloudZero](https://www.cloudzero.com/blog/claude-opus-4-8-pricing/)