Anthropic

Claude Sonnet 4.6

Multimodal
New
#1GDPVal AA ELO
#1Finance Agent
#2TAU2-Bench Retail
+7 more

by Anthropic

+
+
+
+
About

Claude Sonnet 4.6 is a general-purpose language model from Anthropic, released in February 2026 as an update to the Sonnet 4 line that introduced adaptive thinking — a mode where the model automatically calibrates its reasoning depth based on task complexity rather than requiring manual configuration by the developer. The model accepts text and image inputs and integrates natively with web search and code execution tools, consolidating capabilities that previously required separate toolchain setup into a unified API surface. It became the primary workhorse model in the Claude 4 series for code assistance, agentic pipelines, and retrieval-augmented applications that benefit from built-in web access.

+
+
+
+
Pricing Range
Input (per 1M)$3.00 -$3.00
Output (per 1M)$15.00 -$15.00
Providers3
+
+
+
+
Timeline
ReleasedFeb 17, 2026
Knowledge CutoffAug 1, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown

Overall Performance

16 benchmarks
Average Score
72.0%
Best Score
97.9%
High Performers (80%+)
5

Performance Metrics

Max Context Window
264.0K

Top Categories

Science
89.9%
Knowledge
89.3%
Agents
83.7%
Multimodal
75.0%
Coding
69.3%
+
+
+
+
All Benchmark Results for Claude Sonnet 4.6
Complete list of benchmark scores with detailed information
TAU2-Bench Telecom
Agents
97.90
97.9%
Self-reported
TAU2-Bench Retail
Agents
91.70
91.7%
Self-reported
GPQA Diamond
Science
89.90
89.9%
Self-reported
MMMLU
Knowledge
89.30
89.3%
Self-reported
GDPVal AA ELO
Agents
1633.00
81.7%
Self-reported
SWE Bench Verified
Coding
79.60
79.6%
Unverified
MMMU-Pro with Tools
Multimodal
75.60
75.6%
Self-reported
BrowseComp
Agents
74.70
74.7%
Self-reported
MMMU-Pro
Multimodal
74.50
74.5%
Self-reported
OSWorld
Agents
72.50
72.5%
Self-reported
Showing 1 to 10 of 16 benchmarks