Anthropic Launches Claude Opus 4.8 With Stronger Coding and AI Agent Capabilities

Anthropic Launches Claude Opus 4.8 With Stronger Coding and AI Agent Capabilities

Anthropic has announced the release of Claude Opus 4.8, the latest version of its flagship AI model, bringing improvements in coding, reasoning, agent workflows, and professional knowledge tasks.

The new model is available through Claude.ai, Claude Code, and the Claude API under the name claude-opus-4-8. According to Anthropic, the upgrade delivers better performance while maintaining the same pricing structure as its predecessor.

One of the biggest additions is a new effort control feature that allows users to choose how much computational effort Claude spends on a task. This gives users more control over the balance between response quality, speed, and token costs. The model defaults to a high-effort setting, while an "xhigh" mode is available for more demanding workloads.

Anthropic has also introduced dynamic workflows in Claude Code, enabling the AI to plan tasks, launch multiple sub-agents, verify results, and report findings back to users. The feature is designed for large software projects and can help developers manage codebases containing hundreds of thousands of lines of code.

The company says Claude Opus 4.8 performs better than Claude Opus 4.7 across several benchmark categories, including coding, reasoning, office work, and agent-based tasks. Early testers from industries such as software development, law, finance, and research reported improvements in efficiency and workflow automation.

According to Anthropic, Opus 4.8 is also less likely to approve flawed code without warning. Internal testing found the model was four times less likely to overlook coding errors compared to the previous version. The company also reported lower rates of deceptive or unsafe behavior.

Another notable update is support for live instruction changes through the Messages API. Developers can now adjust permissions, token budgets, or contextual information while an AI agent is actively working, without interrupting the task or restarting the conversation.

Anthropic says these improvements are part of a broader roadmap focused on more powerful and cost-efficient AI systems. The company also revealed progress on Project Glasswing, a cybersecurity initiative that uses its experimental Mythos-class models to identify software vulnerabilities and security risks.

The announcement suggests Anthropic is preparing to release even more advanced AI models in the near future, with future systems expected to exceed the capabilities of the current Opus platform while offering improved efficiency and stronger safety protections.