Anthropic

Claude Sonnet 4.5

Multimodal
#2Ï„-bench

by Anthropic

+
+
+
+
About

Claude Sonnet 4.5, released by Anthropic in September 2025, is a large language model from the Claude 4.5 family that balances response quality and efficiency for coding, agentic tasks, and analytical work. It features a 200K token context window (extendable to 1M tokens in beta), 64K maximum output tokens, native image understanding, and extended thinking support. Sonnet 4.5 targets use cases that require a balance of throughput and reasoning depth, including code generation, data analysis, and multi-step agentic pipelines.

+
+
+
+
Pricing Range
Input (per 1M)$3.00 -$3.00
Output (per 1M)$15.00 -$15.00
Providers3
+
+
+
+
Timeline
ReleasedSep 29, 2025
Knowledge CutoffJan 1, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown

Overall Performance

20 benchmarks
Average Score
56.2%
Best Score
98.0%
High Performers (80%+)
5

Performance Metrics

Max Context Window
264.0K

Top Categories

Knowledge
89.5%
Science
83.4%
Agents
68.6%
Coding
58.4%
Finance
54.5%
+
+
+
+
All Benchmark Results for Claude Sonnet 4.5
Complete list of benchmark scores with detailed information
TAU2-Bench Telecom
Agents
98.00
98.0%
Self-reported
MMMLU
Knowledge
89.50
89.5%
Self-reported
TAU2-Bench Retail
Agents
86.20
86.2%
Self-reported
Ï„-bench
Agents
84.70
84.7%
Unverified
GPQA Diamond
Science
83.40
83.4%
Unverified
SWE Bench Verified
Coding
77.20
77.2%
Self-reported
MMMU-Pro with Tools
Multimodal
68.90
68.9%
Self-reported
GDPVal AA ELO
Agents
1276.00
63.8%
Self-reported
MMMU-Pro
Multimodal
63.40
63.4%
Self-reported
OSWorld
Agents
61.40
61.4%
Unverified
Showing 1 to 10 of 20 benchmarks