Claude Sonnet 5.5
ClaudeBalanced
ThinkingTool UseVisionStructured Output
About this model
Claude Sonnet 5.5 is Anthropic's Sonnet-class model for well-scoped everyday work, succeeding Claude Sonnet 5 as a direct upgrade. It is especially strong at building features, fixing bugs, and producing...
Performance Tier
Balanced
Claude Sonnet 5.5 is a balanced model from Claude : strong performance at a reasonable price.
Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.
Pricing
This model is included in Elosia plansHigh
High pricing. Reserve this level for tasks that demand maximum quality.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $2.00 |
| Output (completion) | $10.00 |
| Cache read | $0.100 |
| Cache write | $2.50 |
Capabilities
Context Length1.0M
Max Output Tokens128K
TokenizerClaude
Inputtext, image, file
Outputtext
Release DateSeptember 28, 2026
Benchmarks
General Intelligence
MMLU
Not reported
Mathematics
MATH-500
Not reported
Programming
HumanEval
Not reported
SWE-bench Multilingual
90.3%
Reasoning
Humanity's Last Exam
56.9%
Agentic
SWE-bench Pro
81.3%
Where does Claude Sonnet 5.5 stand?
Compare its performance index against every other model.
Recommended Use Cases
CodingAnalysisResearchGeneral Chat
Strengths
- Agentic coding and bug fixing on well-scoped tasks, with quick understanding of a codebase
- Polished documents, spreadsheets and slides, including decks built from a provided slide template
- Adaptive thinking on by default
- 1M-token context window, 128K output, and text, image and file input, with strong image understanding, including charts and screenshots
- Anthropic's testing shows it rarely over-refuses benign requests, and it is available with zero data retention
Limitations
- Its cybersecurity safeguards can refuse cybersecurity tasks, including benign ones, and some microbiology and virology requests may be flagged in error
- Anthropic publishes no GPQA Diamond, SWE-bench Verified, MMMU-Pro or ARC-AGI-2 score for this model
Frequently asked questions
Resources
This model may use your data for training