Anthropic has launched Claude Opus 4.8, the latest version of its flagship AI model. The upgrade builds on Opus 4.7 with improvements across multiple benchmarks and enhanced collaboration capabilities, while keeping the same pricing structure.
The release comes with several new features designed to give users more control over their AI interactions. Users can now adjust how much effort Claude puts into tasks, while a new “dynamic workflows” feature in Claude Code allows the model to tackle large-scale programming problems using hundreds of parallel subagents.
Performance improvements across the board
Early testers report that Claude Opus 4.8 shows better judgment and reliability when performing complex tasks. The model demonstrates improved performance on coding, reasoning, and practical knowledge work compared to its predecessor and competing models.
Key improvements include:
- Better tool calling efficiency with fewer steps needed for the same intelligence level
- Improved end-to-end task completion rates
- Higher scores on coding benchmarks and agent evaluations
- Enhanced multimodal capabilities for processing PDFs and diagrams
One significant upgrade is the model’s honesty. Opus 4.8 is approximately four times less likely than its predecessor to let flawed code pass without comment. The model is also more likely to flag uncertainties in its work rather than making unsupported claims.
New features roll out today
Alongside the model upgrade, Anthropic is introducing several new capabilities. The effort control feature allows users to choose how much computational power Claude applies to their requests. Higher effort settings produce better responses but use more tokens, while lower settings provide faster responses and preserve rate limits.
Dynamic workflows in Claude Code represents a major capability expansion. This feature can handle codebase-scale migrations across hundreds of thousands of lines of code, using the existing test suite as a quality benchmark. The system plans work, runs parallel subagents, and verifies outputs before reporting back to users.
The Messages API also gets an update, allowing developers to add system entries within the messages array. This lets them update Claude’s instructions mid-task without breaking the prompt cache.
Pricing and availability
Claude Opus 4.8 maintains the same pricing as its predecessor: $5 per million input tokens and $25 per million output tokens for regular usage. Fast mode, which operates at 2.5 times normal speed, costs $10 per million input tokens and $50 per million output tokens – three times cheaper than previous models.
The model defaults to high effort settings, which Anthropic says provides the best balance of quality and user experience. For coding tasks, this effort level uses similar token counts to Opus 4.7’s default while delivering better performance.
Anthropic has also increased rate limits in Claude Code to accommodate higher token usage at elevated effort levels, giving users flexibility to choose settings that match their project needs.
What’s coming next
While Opus 4.8 represents a solid incremental improvement, Anthropic has bigger plans ahead. The company is working on models that provide similar capabilities to Opus at lower costs, addressing one of the key barriers to widespread AI adoption.
More significantly, Anthropic plans to release a new class of models with higher intelligence than current Opus versions. As part of “Project Glasswing,” a limited number of organizations are already testing Claude Mythos Preview for cybersecurity work. These more capable models require stronger safety measures before general release, but Anthropic expects to make Mythos-class models available to all customers within the coming weeks.
This roadmap reflects the broader AI industry’s push toward both more capable and more accessible models. As companies compete to build the most powerful AI systems, they’re also working to make these tools practical for everyday business use through better pricing and user controls.
