Robots Atlas>ROBOTS ATLAS
Claude Sonnet 4.5

Claude Sonnet 4.5

4.5ย ยทย Family: Claude
Anthropic's Claude family AI model focused on coding, agentic tasks and computer use, with a 200K-token context window.
โœ“ Activeโœ“ Public accessLLMMultimodal๐Ÿ“ Claude
Context window
200K
tokens
Max output
64,000
tokens
Release date
29 September 2025
Access:APIHostedDeployment:โ˜ Cloud

Overview

Claude Sonnet 4.5 is a Claude family language model released by Anthropic on September 29, 2025. Anthropic positions it as a leading model for coding, building complex agents, and computer use, with substantial gains in reasoning and math compared to Sonnet 4. The model accepts text and image input and produces text output, including code. The default context window is 200,000 tokens and maximum output is 64,000 tokens. The API identifier is claude-sonnet-4-5 (snapshot claude-sonnet-4-5-20250929). It is available via the Claude API, AWS Bedrock, Google Vertex AI, and Microsoft Foundry. Pricing is USD 3 per 1M input tokens and USD 15 per 1M output tokens. Sonnet 4.5 supports extended thinking. Anthropic deployed the model under AI Safety Level 3 (ASL-3) safeguards.

Classification
LLMMultimodal
Family: Claude
Access & deployment
APIHosted
Cloud
Weights: Closed
Key parameters
๐Ÿ“ Context: 200K
โœ“ Tools
๐Ÿ“ฅ Input: text, image

Technical specification

Context window
200K
tokens
Max output tokens
64,000
tokens per response
Features:โœ“ Tool use
Modalities
โฌ‡ Input
textimage
โฌ† Output
textcode

Capabilities and applications

Native model capabilities
Coding
Category: coding
Reasoning
Category: reasoning
Multi-step reasoning
Category: reasoning
Long context
Category: reasoning
Multilingual
Category: language
Image understanding
Category: vision
Function Calling
Category: planning
Parallel Tool Calls
Ability to invoke multiple external tools simultaneously while generating a response.
Category: reasoning
Planning
Category: planning
Agentic capability
The model's ability to autonomously plan and execute multi-step tasks by sequentially using tools, maintaining context, and adapting to intermediate results.
Category: planning
Computer use
The model's ability to operate a computer interface by interpreting screenshots and generating actions such as clicks, typing, and navigating applications.
Category: planning

Benchmark results

2 benchmarks
SWE-bench
accuracy ยท SWE-bench Verified, 200K thinking budget, averaged over 10 trials
77.2%
๐Ÿ“… 29 Sept 2025๐Ÿ“„ Anthropic announcement (Introducing Claude Sonnet 4.5)
OSWorld
accuracy ยท OSWorld-Verified framework, 100 max steps, averaged across 4 runs
61.4%
๐Ÿ“… 29 Sept 2025๐Ÿ“„ Anthropic announcement (Introducing Claude Sonnet 4.5)

Pricing

Technical architecture