Claude Sonnet 4.5

Summary

Claude Sonnet 4.5 is Anthropic's balanced flagship model, released September 29, 2025. It sits in the sweet spot between Haiku's speed and Opus's capability — delivering state-of-the-art performance on real-world agentic tasks, coding, and computer use at a price point accessible for production deployment.

Overview

Claude Sonnet 4.5 is Anthropic's balanced flagship model, released September 29, 2025. It sits in the sweet spot between Haiku's speed and Opus's capability — delivering state-of-the-art performance on real-world agentic tasks, coding, and computer use at a price point accessible for production deployment. Sonnet 4.5 is the model most developers reach for first, and for good reason: it ranked as the top model on SWE-bench Verified at launch with a score of 77.2%, and it can sustain autonomous operation for over 30 hours on complex multi-step tasks.

What distinguishes Sonnet 4.5 from earlier models is its reliability in long-running agentic contexts. Where previous models would lose coherence or make costly errors over extended autonomous sessions, Sonnet 4.5 was explicitly engineered for sustained agent performance — making it the go-to choice for coding agents, computer use workflows, and multi-step automation pipelines.

Specifications

  • Developer: Anthropic
  • Model String: claude-sonnet-4-5-20250929
  • Release Date: September 29, 2025
  • Type: Large Language Model (LLM), Multimodal (text + vision)
  • Context Window: 200,000 tokens
  • Max Output: 8,192 tokens
  • Access: Anthropic API, Amazon Bedrock, Google Cloud Vertex AI, Claude.ai (Free and Pro)
  • Pricing: $3 per million input tokens / $15 per million output tokens

Capabilities

Coding & Software Engineering: Scored 77.2% on SWE-bench Verified at launch, the highest of any publicly available model at that time. Handles complex multi-file edits, debugging, and code generation across a wide range of languages and frameworks.

Agentic Tasks: Designed specifically for long-horizon autonomous operation. Can maintain consistent behavior and goal-tracking over 30+ hour sessions, making it reliable for complex workflows that require sustained reasoning and tool use.

Computer Use: Strong performance on computer use tasks — navigating interfaces, filling forms, extracting data from visual inputs, and interacting with applications programmatically.

Vision: Handles image analysis, chart interpretation, and visual reasoning tasks effectively alongside text.

Tool Use: Native support for function calling and tool use, enabling tight integration with external APIs, databases, and services.

Limitations

For the hardest reasoning tasks — complex mathematical proofs, cutting-edge research synthesis, or tasks requiring the maximum depth of analysis — Opus 4.5 or 4.6 will outperform Sonnet. Sonnet 4.5 also has a fixed 200K context window, compared to Opus/Sonnet 4.6's optional 1M context via beta header.

Recent Developments

  • September 2025 Launch: Released as the best model for real-world agents and coding at the time, immediately displacing earlier Claude versions in most production use cases.
  • Developer Adoption: Quickly became the most widely used Claude model in API integrations due to its price/performance ratio and agentic reliability.
  • Succeeded by Sonnet 4.6: As of February 17, 2026, Claude Sonnet 4.6 is the recommended default for most use cases, with Sonnet 4.5 remaining available for developers who prefer it.

Last Updated

February 26, 2026

→ Back to Models