Claude Opus 5 is Anthropic’s newest and strongest Opus model, built for advanced coding, professional analysis and long-running AI-agent tasks. Anthropic released the model on July 24, 2026, at the same standard API price as Claude Opus 4.8: $5 per million input tokens and $25 per million output tokens.
Opus 5 is now the default model for Claude Max subscribers and the strongest model available through Claude Pro. It is also available through the Claude API, major cloud platforms and GitHub Copilot.
What Is Claude Opus 5?
Claude Opus 5 is a premium reasoning model in Anthropic’s Opus family. It is designed for difficult tasks that require careful planning, repeated tool use and verification across several connected steps.
The model is particularly suited to complex software development, large codebase changes, document-heavy professional work, spreadsheets, presentations and AI agents that must continue working for extended periods.
Anthropic describes Opus 5 as a step-change improvement over Opus 4.8 rather than a small incremental upgrade. Its biggest reported gains appear in deep reasoning, agentic coding, long-horizon work, code review, visual understanding and multi-agent coordination.
Opus 5 is not Anthropic’s most powerful model overall. Claude Fable 5 remains the company’s most capable widely released model. However, Opus 5 approaches Fable 5 on several coding and professional-work evaluations while costing half as much through the standard API.
How Much Does Claude Opus 5 Cost?
Claude Opus 5 uses the following standard API pricing:
| Usage type | Price |
|---|---|
| Input tokens | $5 per million tokens |
| Output tokens | $25 per million tokens |
| Prompt-cache write | $6.25 per million tokens |
| Prompt-cache read | $0.50 per million tokens |
| Fast mode input | $10 per million tokens |
| Fast mode output | $50 per million tokens |
The standard token price is unchanged from Claude Opus 4.8. Anthropic also offers batch processing, which can reduce eligible API processing costs by 50 percent, and prompt caching, which can reduce the cost of repeatedly processing the same long prompt.
Fast mode is a separate research-preview option that runs Opus 5 at approximately 2.5 times the default speed. It costs twice the standard API rate.
Fast mode is currently available only through the Claude API. It is not currently available through Amazon Bedrock, Google Cloud or Microsoft Foundry.
These API charges are separate from Claude Pro, Max, Team and Enterprise subscription fees.
Is Claude Opus 5 Included With Claude Pro and Max?
Claude Opus 5 is available to Claude Pro, Max, Team and Enterprise users.
Anthropic identifies Opus 5 as the strongest model available through Claude Pro. It is the default model for Claude Max subscribers, although actual use remains subject to each account’s plan limits and model-usage allowances.
Claude Pro currently costs $20 per month when billed monthly in the United States. An annual subscription costs $200 upfront, which works out to approximately $17 per month.
Claude Max starts at $100 per month and provides either five or 20 times more usage than Claude Pro, depending on the selected Max option. Listed prices do not include applicable taxes and can change.
A Claude subscription does not automatically provide API credits. Developers using claude-opus-5 through the Claude Platform receive a separate usage bill based on input and output tokens.
What Are the Claude Opus 5 Specifications?
The main Claude Opus 5 API specifications are:
| Specification | Claude Opus 5 |
|---|---|
| API model ID | claude-opus-5 |
| API context window | 1 million tokens |
| Maximum output | 128,000 tokens |
| Default reasoning setting | Thinking enabled |
| Effort levels | Low, medium, high, xhigh and max |
| Standard input price | $5 per million tokens |
| Standard output price | $25 per million tokens |
| Fast mode | API only, at twice the standard price |
The one-million-token context window is both the default and maximum API context for Opus 5. Anthropic does not offer a smaller-context API version of the model.
Opus 5 can generate up to 128,000 output tokens in one request, although the actual response size depends on the configured output limit and the model’s reasoning requirements.
The one-million-token specification should not be confused with the context shown for individual Claude subscriptions. Anthropic’s current consumer-plan comparison lists a 200,000-token context window across its individual Free, Pro and Max options.
Therefore, a Claude Pro subscriber should not assume every normal Claude chat automatically receives the full one-million-token API context window.
What Are the Main Claude Opus 5 Features?
Stronger Agentic Coding
Agentic coding is one of the model’s most important improvements.
Opus 5 can work across several files, investigate difficult bugs, review existing architecture and continue through longer implementation tasks. Anthropic says the model is less likely than previous Opus versions to leave unfinished placeholders or partially implemented sections when handling larger assignments.
It can also use tools, inspect its progress and revise its work before completing a task. Anthropic’s launch examples include the model identifying a software bug’s underlying cause after another model corrected only the visible symptom.
On Anthropic’s reported Frontier-Bench results, Opus 5 more than doubled Opus 4.8’s performance while achieving a lower average cost per completed task. These are vendor-reported benchmark results, so development teams should still test the model against their own codebases and workflows.
Better Verification and Error Correction
Opus 5 checks its own work more actively than earlier Opus models.
The model may run tests, examine results and make corrections without being explicitly instructed to verify every step. Anthropic’s developer guidance warns that older prompts containing repeated verification instructions can cause Opus 5 to spend too much time checking the same work.
This behaviour can improve coding, interface testing, spreadsheet analysis and other tasks where a plausible-looking answer may still contain functional mistakes. However, human review remains necessary for production code, financial calculations, legal documents and other high-impact decisions.
One-Million-Token API Context
The one-million-token context window allows Opus 5 to process large amounts of information in a single API conversation.
Developers can use this capacity for extensive code repositories, large collections of documents, lengthy research material or complicated agent sessions involving many tool results.
Anthropic says the model maintains instruction-following, reasoning and tool-calling consistency throughout the context window. Actual performance can still depend on how clearly the information is organised and whether the most important instructions are easy for the model to identify.
Adjustable Effort Levels
Opus 5 supports five effort levels:
- Low
- Medium
- High
- Xhigh
- Max
Effort controls how much computational work the model applies to a request. Lower settings can reduce token use and response time, while xhigh and max allow deeper analysis for difficult tasks.
Anthropic recommends testing different effort settings instead of automatically using max for every request. Low and medium effort can produce strong results on less demanding work at a fraction of the token use and latency.
Thinking is enabled by default. Developers can disable it only at high effort or below. Attempting to disable thinking while using xhigh or max returns an API error.
Improved Professional Work
Claude Opus 5 also targets work outside software development.
Anthropic and its launch partners report improvements in numerical reasoning, table analysis, financial research, due diligence, legal document review, spreadsheets and presentation creation.
The model can work with complex multi-sheet spreadsheets containing formulas and create more structured slide decks. It may also challenge a questionable assumption rather than automatically following an unsuitable proposed approach.
These capabilities do not guarantee that every generated calculation, contract edit or business conclusion will be correct. Important work still requires review by someone with relevant subject knowledge.
Stronger Visual and Computer-Use Performance
Opus 5 can analyse charts, diagrams, documents and visual interfaces. Anthropic says the model is particularly capable when it can repeatedly inspect, crop and verify visual material using tools.
The company also demonstrated Opus 5 creating interactive visual outputs, including a simplified cell illustration and a wind-tunnel simulation.
On Anthropic’s reported OSWorld 2.0 computer-use results, Opus 5 surpassed Fable 5’s best result at slightly more than one-third of the cost per task. This does not mean Opus 5 is universally better than Fable 5, but it suggests that Opus 5 may offer a stronger cost-to-performance balance for some computer-use workflows.
How Is Claude Opus 5 Different From Opus 4.8?
Claude Opus 5 keeps Opus 4.8’s standard API price but changes several capabilities and default behaviours.
The most important differences include:
- A one-million-token context window by default
- Thinking enabled automatically
- A new max effort setting
- Stronger long-running coding performance
- Better code review and bug finding
- More active self-verification
- Stronger visual understanding
- Better multi-agent coordination
- Improved spreadsheets and presentation creation
- A lower minimum length for prompt caching
The minimum cacheable prompt length has fallen from 1,024 tokens on Opus 4.8 to 512 tokens on Opus 5. This allows developers to cache shorter recurring prompts without changing their application code.
Developers migrating an existing application must update the model ID from claude-opus-4-8 to claude-opus-5. They should also review output limits and thinking settings because Opus 5 enables thinking by default.
Claude Opus 4.8 remains available, so developers do not need to migrate every production workload immediately.
Is Claude Opus 5 Better Than Claude Fable 5?
Claude Opus 5 is not better than Fable 5 in every area.
Fable 5 remains Anthropic’s most capable widely released model and is intended for the company’s most demanding long-running tasks. It costs $10 per million input tokens and $50 per million output tokens, exactly twice the standard Opus 5 rates.
However, Opus 5 comes close to Fable 5 on several reported coding evaluations.
At maximum effort, Opus 5 performed within 0.5 percentage points of Fable 5’s highest CursorBench score while costing half as much per task. Opus 5 also reportedly surpassed Fable 5 on OSWorld 2.0 at a much lower task cost.
The practical difference is value rather than absolute superiority.
Fable 5 remains appropriate when an organisation needs Anthropic’s highest available capability and can justify the higher price. Opus 5 may be the more economical option for developers who want near-frontier coding and agent performance for regular production work.
Where Is Claude Opus 5 Available?
Claude Opus 5 is available through:
- Claude Pro
- Claude Max
- Claude Team
- Claude Enterprise
- Claude API
- Claude Code
- Claude Cowork
- Amazon Bedrock
- Google Cloud
- Microsoft Foundry
The Claude API model ID is claude-opus-5. On Amazon Bedrock, Anthropic documents the model as anthropic.claude-opus-5.
India appears on Anthropic’s current list of supported countries for both Claude.ai and commercial API access. Indian users can therefore access supported Claude services, subject to account eligibility, plan limits, billing requirements and product availability.
Is Claude Opus 5 Available in GitHub Copilot?
Claude Opus 5 is available in GitHub Copilot for eligible subscribers.
GitHub is gradually rolling it out to Copilot Pro+, Max, Business and Enterprise users. It can be selected through supported Copilot model pickers in:
- Visual Studio Code
- Visual Studio
- Copilot CLI
- GitHub Copilot cloud agent
- GitHub Copilot app
- GitHub.com
- GitHub Mobile
- JetBrains
- Xcode
- Eclipse
Copilot Business and Enterprise administrators must enable the Claude Opus 5 policy before organisation members can use it.
GitHub bills Opus 5 at the provider’s API list price under usage-based billing. Because the rollout is gradual, some eligible users may not see the model immediately.
What Safety Restrictions Does Opus 5 Have?
Opus 5 includes stronger safeguards for a narrow group of cybersecurity requests.
Anthropic says the model can identify vulnerabilities in source code, but its safety classifiers may block binary-based vulnerability scanning, penetration testing and exploit generation.
When a request triggers these classifiers in Claude.ai, Claude Code or Claude Cowork, the system can route it to Claude Opus 4.8 instead. Developers can also enable automatic fallbacks through the API.
Anthropic reports that Opus 5 remains behind Mythos 5 in offensive cybersecurity and long-running autonomous biological research. The company therefore does not consider Opus 5 an advancement of the frontier in those higher-risk areas.
Is Claude Opus 5 Worth Using?
Claude Opus 5 is a meaningful upgrade for developers and professionals who already rely on Opus models for demanding work.
Its strongest advantages are the one-million-token API context, 128,000-token maximum output, improved agentic coding, deeper effort controls and more reliable self-verification. It also keeps the same standard API price as Opus 4.8.
Opus 5 will not be the most economical option for every request. Claude Sonnet may provide better value for routine writing, simple coding and high-volume tasks. Fable 5 remains more capable for some exceptionally difficult workloads but costs twice as much through the standard API.
For users who need high-end coding, analysis and AI-agent capabilities without paying Fable 5 rates, Claude Opus 5 offers Anthropic’s strongest current balance of capability, availability and cost.
Frequently Asked Questions
How much does Claude Opus 5 cost?
Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens through the standard Claude API. Its faster API mode costs $10 per million input tokens and $50 per million output tokens.
Does Claude Opus 5 have a one-million-token context window?
Yes, Claude Opus 5 supports a one-million-token context window through the Claude API. Normal Claude subscription chats may have different context limits, so users should not assume every Pro or Max conversation automatically receives the full API capacity.
What is the maximum output length of Claude Opus 5?
Claude Opus 5 supports a maximum output of 128,000 tokens through the API. The actual output depends on the developer’s configuration, available context and the complexity of the request.
Is Claude Opus 5 better than Fable 5?
Claude Opus 5 is not universally better than Fable 5. Fable 5 remains Anthropic’s highest-capability option for exceptionally demanding work, but Opus 5 approaches it on several reported evaluations while costing half as much through the standard API.
Is Claude Opus 5 available in GitHub Copilot?
Yes. GitHub is rolling out Claude Opus 5 to eligible Copilot Pro+, Max, Business and Enterprise subscribers. Business and Enterprise administrators may need to enable the model before organisation members can select it.
