Anthropic has introduced Stop Task 4.8An update to Cloud Opus 4.7 which the company claims improves performance for coding, agentic tasks, reasoning, and knowledge work.
The model is accessible via claude.ai, Cloud Code, and Cloud API, where it is identified as cloud-opus-4-8. The release also includes new platform controls, including user-adjustable effort settings, dynamic workflows in Cloud Code, and live updates of the messaging API.
Pricing and Effort Control in Cloud Opus 4.8
Cloud Opus 4.8 maintains the same pricing structure as Cloud Opus 4.7. Standard mode costs five dollars per million input tokens and twenty-five dollars per million output tokens.
Fast Mode costs ten dollars per million input tokens and fifty dollars per million output tokens. Additionally, the fast mode for Opus 4.8 operates at 2.5 times the speed of the standard mode.
Users on claude.ai and Cowork can now adjust how much effort Cloud puts into responding, which directly impacts token usage. Opus 4.8 defaults to the high effort setting.
Anthropic says that even at this default level, coding tasks use the same tokens as in Opus 4.7, while providing better performance.
For tasks that require more computation, the xhigh setting is available. The cloud’s code rate limits have been increased to support the higher token consumption enabled by these effort controls.
These new options are part of Anthropic’s shift toward token-based billing rather than fixed subscription tiers.
Live updates of dynamic workflow and messaging APIs in Cloud Code
Cloud Code now supports dynamic workflows, enabling users to plan tasks, run sub-agents in parallel, verify output, and report results.
Anthropic explains that this feature is targeted for larger codebases and can handle codebases containing hundreds of thousands of lines. Currently in Research Preview, Dynamic Workflows are available for Enterprise, Team, and Max plans.
The messaging API now supports real-time updates to the message array when the agent is active. Developers can change instructions during a task, adjust permissions, modify the token limit, or update the context without disrupting instant caching or requiring a new user.
Performance and security changes in Cloud Opus 4.8
Anthropic claims that Opus 4.8 offers improvements over Opus 4.7 in various benchmarks, including coding, agentic functions, reasoning, and office tasks.
The company also says that this model is almost four times less likely to pass an erroneous code compared to the Opus 4.7, with no comment on this.
In terms of security, Anthropic reports that Opus 4.8 shows lower levels of deception and less propensity to comply with attempts to misuse the model than Opus 4.7. These security metrics are said to be similar to those in the Cloud Mythos Preview.
External testers cited by Anthropic include organizations involved in software development, law, finance, and research. CursorBench found that Opus 4.8 required fewer tool steps to achieve the same output quality as comparable models.
Another tester noted that the model’s cost performance was comparable to GPT-5.5 in internal benchmarks.
looking ahead
Anthropic said it is developing models that offer existing capabilities at a lower cost, while also introducing a new class of models that surpasses the existing Opus level.
The Mythos-class models, which are currently limited to participants in Project Glasswing, are expected to be available to customers in the next few weeks after additional safety measures are implemented.
Cloud Opus 4.8 is now accessible to users through claude.ai, Cloud Code, and Cloud API.
