Claude Opus 4.8 Explained

A plain-language guide to Anthropic's value-tier model: what it does, what it costs, the compliance rules, and when to route work to it instead of Fable 5.

Claude Opus 4.8 is Anthropic's general-purpose coding and agentic model, released May 28, 2026, at $5 per million input tokens and $25 per million output tokens (Anthropic). This guide explains Claude Opus 4.8 in plain terms. It covers the model's features, price, best business uses, compliance rules, and how it compares.

Opus 4.8 is the value tier for routine coding and agentic work. It powers Claude Code and runs at about half the price of Claude Fable 5. That makes it the model to route most day-to-day work to, and to save Fable 5 for the hardest problems.

Below we break down each part so you can decide when Opus 4.8 fits. We use Anthropic's own sources for every model claim.

Reviewed by Jonathan West, Founder of Layer3 Labs, on July 5, 2026. We research using primary vendor and regulator sources.

What Is Claude Opus 4.8?

Claude Opus 4.8 is Anthropic's strong general-purpose coder and agentic-workflow model, released May 28, 2026 (Anthropic). It writes and reviews code, runs multi-step tasks, and powers Claude Code.

It sits below Claude Fable 5 in Anthropic's lineup, but it costs about half as much. That price gap is the whole point of the model.

Think of Opus 4.8 as your default work model. Route routine coding and agentic jobs here, then reserve Fable 5 for the hardest problems.

  • Released May 28, 2026 by Anthropic.
  • Strong general-purpose coder across many files.
  • Built for agentic, multi-step workflows.
  • Powers Claude Code for developer tasks.
  • About half the price of Claude Fable 5.
Sources
Anthropic — Models overview

Deciding when to route work to Claude Opus 4.8 versus Fable 5? We can map the right tier to your workflows and cut your model spend on routine coding.

Book a Consultation

Key Features and Capabilities

Claude Opus 4.8's main strength is reliable coding and agentic work at a lower price (Anthropic). It handles the bulk of real development tasks without the top-tier cost.

The model is built to act, not just answer. It runs multi-step jobs, calls tools, and works through a task with less hand-holding.

It also powers Claude Code. That means you get a capable coding model inside a developer workflow you already use.

  • Strong general-purpose software engineering.
  • Agentic workflows: plans and executes multi-step tasks.
  • Tool use for automation across steps.
  • Powers Claude Code for developer workflows.
  • Standard Claude safety guardrails built in.

Pricing and Access

Claude Opus 4.8 costs $5 per million input tokens and $25 per million output tokens (Anthropic). Input is what you send in. Output is what the model writes back.

That is about half the price of Claude Fable 5, which runs at $10 input and $50 output. For the same volume of routine coding, Opus 4.8 cuts your model bill roughly in half.

You reach Opus 4.8 through Anthropic's own products, including Claude Code. Output tokens cost five times more than input, so concise prompts and outputs save money.

  • Input: $5 per million tokens.
  • Output: $25 per million tokens.
  • About half the price of Claude Fable 5.
  • Access: Anthropic products, including Claude Code.
  • Output costs five times input, so keep prompts tight.

Best Uses for Business

Claude Opus 4.8 fits routine, high-volume coding and agentic work best. Its lower price makes it the model to run for the everyday jobs that fill a workweek.

The smart pattern is a two-tier setup. Send routine work to Opus 4.8, and escalate only the hardest problems to Fable 5. That keeps quality high and cost low.

  • Software teams: day-to-day coding, review, and bug fixes in Claude Code.
  • Automation: agentic workflows that call tools and run multi-step tasks.
  • Internal tooling: scripts, data cleanup, and back-office code.
  • Drafting and analysis: routine knowledge work at volume.
  • Cost control: default tier that saves Fable 5 for the hardest problems.
Decision point: route routine coding to Opus 4.8 and escalate only the hardest problems to Fable 5. This two-tier split roughly halves your model spend on everyday work.

How Claude Opus 4.8 Compares

Claude Opus 4.8 is the value tier: about half the price of Fable 5 and above Sonnet 4.6 for agentic coding (Anthropic). Its main peers are Claude Fable 5 and Claude Sonnet 4.6.

Against Fable 5, Opus 4.8 is the value pick for routine coding. Fable 5 is Anthropic's most capable public model at $10 input and $50 output, so save it for the hardest problems. Opus 4.8 also serves as the safety fallback Fable 5 hands off to on blocked requests.

Against Sonnet 4.6, Opus 4.8 is the stronger agentic coder. Sonnet 4.6 is cheaper at $3 input and $15 output, with a 1M-token context window in beta, and suits long-context and high-throughput jobs. Use the table to match a model to your need.

  • vs Fable 5: Opus 4.8 is about half the price and the value pick for routine coding; Fable 5 is the top model for the hardest problems.
  • vs Fable 5: Opus 4.8 is also the safety fallback Fable 5 uses when a request is blocked.
  • vs Sonnet 4.6: Opus 4.8 is the stronger agentic coder; Sonnet 4.6 is cheaper with a 1M-token context window for long-context work.
Comparison table: Opus 4.8 vs Fable 5 vs Sonnet 4.6. | Price (input/output per M tokens): Opus 4.8 = $5/$25; Fable 5 = $10/$50; Sonnet 4.6 = $3/$15. | Role: Opus 4.8 = value default for routine coding and agentic work; Fable 5 = top model for the hardest problems; Sonnet 4.6 = cheapest, best for long-context and high-throughput. | Context: Sonnet 4.6 = 1M-token window (beta). | Verdict: route routine coding to Opus 4.8, escalate the hardest problems to Fable 5, and use Sonnet 4.6 for long documents on a budget.
Sources
Anthropic — Models overview

Limitations and Risk Factors

Claude Opus 4.8's main limit is that it is not Anthropic's most capable model (Anthropic). For the hardest problems, Fable 5 will outperform it.

It carries standard Claude safety guardrails. Safeguards can decline high-risk requests, so plan for that in sensitive workflows.

Like any model, its output needs review. It can be wrong or incomplete, so keep a human in the loop for anything that ships or touches regulated data.

  • Not the top model; Fable 5 wins on the hardest tasks.
  • Standard Claude guardrails can decline high-risk requests.
  • Output can be wrong or incomplete.
  • Human review is still required for every output.
  • Escalate the hardest problems to Fable 5 rather than pushing Opus 4.8 past its tier.

Compliance and Data Handling

Claude Opus 4.8 inherits Anthropic's SOC 2, ISO 27001, and ISO 42001 certifications, with a HIPAA BAA available on the API and Enterprise plans (Anthropic). These cover the platform, not your own legal duties.

It runs under standard Claude safety guardrails. Unlike the Mythos-class Fable 5, Opus 4.8 does not carry a mandatory 30-day safety retention rule tied to that tier.

For regulated work, confirm the details before you go live. Sign a HIPAA BAA where required and check your own data-handling terms match the model you send data to.

  • Certifications: SOC 2, ISO 27001, ISO 42001.
  • HIPAA BAA available on the API and Enterprise plans.
  • Standard Claude safety guardrails.
  • Not subject to Fable 5's Mythos-class 30-day retention rule.
  • Certifications cover the platform, not your own compliance duties.
Decision point: sign a HIPAA BAA on the API or Enterprise plan before sending regulated data to Claude Opus 4.8, and confirm your own data-handling terms match.
Sources
Anthropic — Trust Center

Frequently Asked Questions

  • Claude Opus 4.8 is Anthropic's general-purpose coding and agentic-workflow model, released May 28, 2026. It powers Claude Code and runs at about half the price of Claude Fable 5, so it is the value tier for routine work.
  • Claude Opus 4.8 costs $5 per million input tokens and $25 per million output tokens (Anthropic). That is about half the price of Claude Fable 5. Output costs five times more than input, so concise prompts save money.
  • Route routine coding and agentic work to Claude Opus 4.8, and save Claude Fable 5 for the hardest problems. Opus 4.8 costs about half as much, so this two-tier split roughly halves your model spend on everyday tasks.
  • Yes. Claude Opus 4.8 powers Claude Code, Anthropic's coding tool. That gives you a strong general-purpose coder inside a developer workflow, at the lower value-tier price.
  • Claude Opus 4.8 is the stronger agentic coder, while Claude Sonnet 4.6 is cheaper at $3 input and $15 output and has a 1M-token context window in beta. Use Opus 4.8 for routine coding and Sonnet 4.6 for long-context, high-throughput jobs.
  • Claude Opus 4.8 offers a HIPAA BAA on the API and Enterprise plans, and inherits SOC 2, ISO 27001, and ISO 42001 (Anthropic). The certifications cover the platform, not your own compliance duties, which still apply.
  • No. The mandatory 30-day safety retention applies to the Mythos-class Claude Fable 5, not to Claude Opus 4.8. Opus 4.8 runs under standard Claude safety guardrails, so confirm your own data-handling terms before regulated use.
  • Claude Opus 4.8 is not Anthropic's most capable model, so Claude Fable 5 outperforms it on the hardest problems. It also carries standard guardrails that can decline high-risk requests, and its output still needs human review.

Put Claude Opus 4.8 to Work in Your Business

Not sure when to route work to Claude Opus 4.8 versus Fable 5? Book a free 30-minute Layer3 Labs AI workflow audit and we will map the right tier to your team.

Book Your Free AI Workflow Audit