Anthropic Unveils Claude Opus 4.8 with Smarter AI Performance and New Features

Anthropic has officially introduced Claude Opus 4.8, the latest version of its flagship AI model. The company says the new release delivers stronger performance across coding, reasoning, and complex task management while remaining available at the same pricing as the previous Opus 4.7 model.

The launch also brings several new tools and workflow improvements designed to enhance productivity for both individual users and developers.

Claude Opus 4.8 Focuses on Better Accuracy and Collaboration

According to Anthropic, Claude Opus 4.8 offers notable improvements in handling coding tasks, agent-based workflows, logical reasoning, and practical knowledge work. The company describes the model as a more capable collaborator that can better assist users with complex projects.

Early testers reported that the new version demonstrates improved judgment and reliability when completing multi-step tasks. They also noted that the model is more transparent about uncertainties and less likely to present unsupported conclusions as facts.

Improved Honesty and Reduced Errors

One of the key upgrades in Claude Opus 4.8 is its focus on accuracy and honesty. Anthropic explained that AI models can sometimes overstate progress or make claims without sufficient evidence.

Testing results indicate that Opus 4.8 is significantly less likely to overlook flaws in generated code compared to its predecessor. The company claims the model is roughly four times better at identifying and acknowledging potential issues during development tasks.

Stronger Safety and Alignment Standards

Before its public release, Claude Opus 4.8 underwent extensive safety evaluations. Anthropic’s Alignment team reported improvements in areas such as user autonomy, helpfulness, and acting in the user’s best interests.

The company also stated that incidents of problematic behavior, including deceptive responses or misuse-related cooperation, were lower than those observed in Claude Opus 4.7. Safety testing results were documented in the model’s official system card.

Dynamic Workflows Arrive in Claude Code

Alongside the new AI model, Anthropic introduced a research-preview feature called Dynamic Workflows for Claude Code.

The feature allows Claude to plan large projects and coordinate hundreds of parallel sub-agents within a single session. These agents can work together on extensive coding tasks, verify outputs, and deliver final results with minimal user intervention.

Anthropic says the technology can support large-scale codebase migrations involving hundreds of thousands of lines of code, making it particularly useful for enterprise development teams.

Users Gain Control Over AI Effort Levels

A new effort control option has also been added to Claude.ai and Cowork. This setting allows users to decide how much processing effort the AI should dedicate to a response.

Higher effort levels provide deeper analysis and potentially better answers, while lower settings prioritize faster response times and reduced resource consumption. The feature is available across all subscription plans.

API Update Simplifies Developer Workflows

Developers using the Claude API will benefit from a new enhancement to the Messages API. The update enables system instructions to be inserted directly into message arrays during active sessions.

This change allows developers to modify permissions, token limits, and environment settings without disrupting existing prompt structures or cache performance.

High-Effort Mode Enabled by Default

Anthropic confirmed that Claude Opus 4.8 operates in high-effort mode by default. The company believes this setting offers the best balance between response quality and efficiency.

Users handling more demanding projects can select “Extra” or “Max” effort levels, enabling the model to dedicate additional computational resources for improved results on complex tasks.

Anthropic Plans More Powerful AI Models

While describing Opus 4.8 as an incremental upgrade, Anthropic revealed plans to develop even more advanced AI systems.

The company is currently testing Claude Mythos Preview through its Project Glasswing initiative, primarily for cybersecurity applications. Anthropic says these next-generation models offer significantly higher intelligence levels but require stronger safety measures before broader public availability.

The company expects Mythos-class AI models to reach more customers in the coming weeks as safety protections continue to improve.

Pricing and Availability

Claude Opus 4.8 is available globally starting today. Pricing remains unchanged from Claude Opus 4.7, with standard usage costing $5 per million input tokens and $25 per million output tokens.

Fast Mode is also available, offering processing speeds up to 2.5 times faster than standard operation. Anthropic says this mode is now significantly more affordable than previous versions, with pricing set at $10 per million input tokens and $50 per million output tokens.

Developers can access the new model through the Claude API using the claude-opus-4-8 model identifier.