Anthropic

Claude Sonnet 4.5

Multimodal
Zero-eval
#1TAU-bench Retail
#1MMMUval
#1SWE-bench Verified (Agentic Coding)
+4 more

by Anthropic

+
+
+
+
About

Claude Sonnet 4.5 was developed to bridge human thinking and machine assistance, allowing people to work with language and tools in a more conversational and natural way. Built with a focus on turning ideas into results, it represents an evolution in making AI feel less mechanical while maintaining strong capabilities across reasoning, coding, and collaborative tasks.

+
+
+
+
Pricing Range
Input (per 1M)$3.00 -$3.00
Output (per 1M)$15.00 -$15.00
Providers2
+
+
+
+
Timeline
AnnouncedSep 29, 2025
ReleasedSep 29, 2025
Knowledge CutoffJan 31, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown

Overall Performance

9 benchmarks
Average Score
75.8%
Best Score
89.1%
High Performers (80%+)
4

Performance Metrics

Max Context Window
264.0K
Avg Throughput
42.0 tok/s
Avg Latency
0ms
+
+
+
+
All Benchmark Results for Claude Sonnet 4.5
Complete list of benchmark scores with detailed information
MMMLU
text
0.89
89.1%
Self-reported
AIME 2025
text
0.87
87.0%
Self-reported
TAU-bench Retail
text
0.86
86.2%
Self-reported
GPQA
text
0.83
83.4%
Self-reported
MMMUval
multimodal
0.78
77.8%
Self-reported
SWE-bench Verified (Agentic Coding)
text
0.77
77.2%
Self-reported
TAU-bench Airline
text
0.70
70.0%
Self-reported
OSWorld
multimodal
0.61
61.4%
Self-reported
Terminal-Bench
text
0.50
50.0%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+