Skip to main content
Sonnet 5 is Anthropic’s next-generation balanced model within DeepMask, carrying the practical intelligence-to-throughput ratio that made the Sonnet 4 series the default for professional engineering into a new generation. It brings Opus-class reasoning and much longer autonomous sessions to the Sonnet tier — so most production agents can run on Sonnet 5 without reaching for Opus.

​
About Sonnet 5

Sonnet 5 (released September 23, 2026) replaces the Standard/High toggle of the Sonnet 4 series with Continuous Reasoning, allocating thinking depth dynamically per step at Sonnet-class cost. It extends autonomous coding to 60+ hour single-objective sessions, raises the OSWorld computer-use score to a new high, and reaches a GPQA Diamond score of 90.1%. Its 2M token context window and improved prompt caching keep high-volume workloads efficient, making it the new default for demanding production deployments.

​
Key Capabilities

Continuous Reasoning

Allocates thinking depth dynamically per step rather than per request — Opus-class reasoning quality at Sonnet-class cost.

60-Hour Autonomous Coding

Manages week-long engineering objectives with self-correction and testing, maintaining coherence across the full session.

Computer Use (Native)

Sees screens, moves cursors, and types in standard desktop applications — enabling true browser-based and GUI automation without external tools.

Cost-Optimized at Scale

Improved prompt caching and batch processing deliver up to 90% cost savings versus standard API calls for high-volume workloads.

​
Best For

Choose Sonnet 5 for enterprise-grade agents, full-stack development lifecycle management, and browser-based automation that need near-Opus accuracy at Sonnet cost — especially long-horizon autonomous engineering and multi-app research workflows. For tasks with lower complexity requirements, Haiku 4.5 offers faster throughput at lower cost. For the most demanding reasoning challenges and large multi-agent systems, Opus 5 sets the ceiling.

​
Use Cases

  • Autonomous software engineering — Building, testing, and deploying full-stack features over multi-day sessions from a single prompt.
  • Complex multi-app workflows — Researching data in a browser and then populating a local Excel sheet and PowerPoint.
  • Enterprise workflow automation — Analyzing financial data, synthesizing internal insights, and generating professional content at scale.
  • Browser-based agents — Automating procurement, competitive analysis, and customer onboarding via digital interaction.
  • Legal and financial forensics — Analyzing massive document sets for subtle logical contradictions.

​
Specifications

Try Sonnet 5 in DeepMask →