Anthropic's Opus 5: Near-Flagship Performance at Half the Cost

30-Second Summary

Anthropic released Claude Opus 5 on July 24, 2026, a new model that delivers performance close to its flagship Claude Fable 5 at roughly half the API price. The model introduces a first-for-Anthropic feature: built-in effort toggles that let users choose between low, medium, and high reasoning intensity for each task, giving businesses a direct lever to manage AI costs without switching models. Opus 5 outperforms Fable 5 on several coding and knowledge benchmarks, carries a knowledge cutoff three months fresher than Fable 5, and is already the default model for Claude Max subscribers.

At a Glance

• Topic: Model Releases
• Company: Anthropic
• Date: July 24, 2026
• Announcement: Anthropic launched Claude Opus 5, a near-flagship model at half the API cost of Fable 5, with built-in effort toggles for cost control.
• What Changed: Claude Opus 5 becomes the default model for Claude Max and the strongest available on Claude Pro, with no price increase for subscribers.
• Why It Matters: Businesses can now access frontier-class AI reasoning at a price point that makes more use cases economically viable.
• Who Should Care: Any organisation using Claude for customer communications, internal research, document analysis, or custom integrations.

Key Facts

• Company: Anthropic
• Launch Date: July 24, 2026
• What Changed: Claude Opus 5 replaces Opus 4.8 as Anthropic's second-tier model, delivering near-Fable-5 performance at $5 per million input tokens and $25 per million output tokens (standard API pricing).
• Who It Affects: Claude Max and Claude Pro subscribers, API integrators, and businesses using Anthropic-powered platforms.
• Primary Sources: TechCrunch (July 24, 2026); Axios (July 24, 2026); Bloomberg (July 24, 2026).

What Happened

Anthropic launched Claude Opus 5 on July 24, 2026, positioning it as an efficient daily driver for business workflows rather than a pure research or capability showcase. The model is priced at $5 per million input tokens and $25 per million output tokens at the standard API tier, making it roughly half the cost of operating Claude Fable 5 at comparable task complexity.

The headline feature is a built-in effort toggle: users and developers can now specify low, medium, or high reasoning effort at the point of the request, allowing simpler tasks such as summarisation, drafting, or data extraction to be handled at lower cost while reserving full compute for complex analysis, multi-step reasoning, or difficult coding problems. Anthropic partners including Harvey (legal AI) and Zapier (automation) have reported cutting token usage significantly since integrating Opus 5 into their workflows.

On benchmark performance, Opus 5 scores 43.3 percent on FrontierBench v0.1, compared to Fable 5's 33.7 percent, outperforming Anthropic's current flagship on coding and structured knowledge tasks. Its training data carries a knowledge cutoff of May 2026, four months fresher than Fable 5's January 2026 cutoff, which has practical implications for businesses relying on AI for research into recent tools, regulations, or market conditions.

Claude Max subscribers gain access to Opus 5 as their default model immediately with no additional charge. Claude Pro subscribers also gain access as the strongest available option on their plan.

Why It Matters

• The effort toggle gives business operators a practical cost lever without requiring model changes or custom prompt engineering. Routing simpler tasks to low-effort mode and complex tasks to high-effort mode can reduce token spend materially across a workflow.
• Pricing at half the API cost of Fable 5 brings use cases that were previously marginal, such as AI-assisted research, first-draft generation at scale, or document review, inside a reasonable budget for smaller organisations.
• The fresher knowledge cutoff (May 2026) reduces the risk of AI outputs based on outdated information, which has been a significant concern for businesses operating in fast-moving regulatory or market environments.
• Opus 5 is already the default for Claude Max and the strongest option on Claude Pro, meaning businesses already paying for those subscriptions receive the upgrade at no additional cost.
• Enterprise partners like Harvey and Zapier are reporting real token cost reductions, providing early evidence that the effort toggle works as intended in production workflows.
• Anthropic's inclusion of smoother API fallback routing and mid-conversation tool swapping lowers the friction of building or maintaining AI-powered systems.

The David and Goliath View

For most business operators running on Claude, the Claude Opus 5 launch is one of those moments where the calculation genuinely changes. The performance curve has been moving in one direction for three years: more capable, and cheaper. Opus 5 is the latest step in that curve, and it is a meaningful one because the effort toggle moves control of cost from the AI provider to the operator.

That matters more than it might sound. Until now, if you wanted to manage your AI bill you had to choose a different model, adjust your prompts, or limit usage. The effort toggle makes this a per-task decision, which means the same model can handle your highest-complexity work and your most routine work without you paying frontier rates for tasks that do not need them. That is how cost discipline and AI capability coexist in a lean operation.

The practical recommendation is straightforward. If you are on Claude Max, Opus 5 is already yours. Spend a few hours this week running your most common AI tasks through it and mapping which effort level fits which type of work. If you have been holding off on expanding AI use because of cost, revisit that decision. The break-even on a number of document-heavy and research-heavy workflows has shifted.

Where This Fits in the AI Stack

AI Growth Engine: Opus 5's improved coding and knowledge performance, combined with lower cost and a fresher knowledge cutoff, makes it the strongest option for businesses building AI-assisted research, content, or sales workflows within an Anthropic stack.

Employee Amplification Systems: The effort toggle is directly suited to employee-facing AI tools. High-effort mode supports complex analysis tasks; low-effort mode handles routine drafting and summarisation without carrying the same cost. Teams can use AI more broadly without budget pressure.

Questions Operators Are Asking

Do I need to do anything to get Claude Opus 5 if I am already on Claude Max? No. Opus 5 is now your default model. You will notice the change the next time you use Claude on the web or via the API if your integration specifies the latest Opus version. If you have hardcoded a specific model version, update your configuration to reference Opus 5.

How does the effort toggle work in practice? When making a request via the Claude API or supported interfaces, you can specify a reasoning effort level: low, medium, or high. Low effort processes the request with less compute and lower cost, which is suitable for simple drafting, summarisation, or formatting tasks. High effort applies deeper reasoning and is better suited to complex analysis, multi-step problem solving, or difficult coding work. Medium sits between the two. Anthropic recommends starting with medium and adjusting based on output quality.

Is Claude Opus 5 better than Claude Fable 5 for all tasks? On coding and structured knowledge benchmarks, Opus 5 outperforms Fable 5. For tasks that depend on creative reasoning, nuanced judgement, or the most complex long-context work, Fable 5 remains Anthropic's flagship. For the majority of business workflows, Opus 5 will deliver equivalent or better results at lower cost.

What does the fresher knowledge cutoff mean for my business? Opus 5's May 2026 knowledge cutoff means it has been trained on more recent information than Fable 5 (January 2026) or Opus 4.8. In practice, this means more accurate responses about recent software versions, regulatory changes, market conditions, and industry developments. For businesses in fast-moving sectors, this reduces the frequency of AI outputs that are confidently wrong about recent facts.

Should this change which AI tools or subscriptions my business uses? If your team currently uses Claude Max or Claude Pro, the upgrade is automatic and there is nothing to change. If you are evaluating AI platforms and comparing cost per quality, Opus 5 strengthens Anthropic's case at the mid-tier pricing band. If your AI spend sits mostly on the API side, review your current model configuration against Opus 5's pricing and determine whether the effort toggle unlocks savings in your highest-volume workflows.

Citable Summary

What happened: Anthropic launched Claude Opus 5 on July 24, 2026, a near-flagship model priced at roughly half the API cost of Claude Fable 5, featuring built-in effort toggles for cost control and a knowledge cutoff of May 2026.

Why it matters: The effort toggle gives business operators a direct lever to manage AI costs by task complexity, making frontier-class reasoning economically viable for a broader range of workflows in organisations of any size.

David and Goliath view: Claude Opus 5 shifts the cost-versus-capability trade-off in favour of smaller organisations. The effort toggle means businesses can access the same top-tier model for both complex and routine work without paying full cost for tasks that do not need it.

Offer relevance:

• AI Growth Engine: direct relevance. Lower cost per output and a fresher knowledge cutoff make AI-powered research, content, and sales workflows more viable.
• Employee Amplification Systems: direct relevance. The effort toggle enables teams to deploy AI broadly across tasks with cost control built in, amplifying employee output without a proportional rise in AI spend.