Back

Anthropic

Claude Sonnet 4.6

ReasoningClosed weights
Intelligence
30.1
Rank 124 of 744 measured models
Coding
63.0
Rank 74 of 298 measured models
Output speed
50 t/s
Rank 319 of 403 measured models
Price/1M
$6
Rank 587 of 659 measured models
Context window
1M
Rank 84 of 1199 measured models
Cost per task
$2.49
Rank 170 of 193 measured models

Claude Sonnet 4.6 (Max) scores 30.1 on the Artificial Analysis Intelligence Index, ranking 124th of 744 measured models. The closest models on this index are DeepSeek V4 Pro 0424 (30.4), GPT-5.2 (30.4), and Claude Sonnet 4.6 (batch) (30.1). At $6 per million tokens (3:1 input/output blended price), it is more expensive than DeepSeek V4 Pro 0424 ($0.544) and GPT-5.2 ($4.81). With an output speed of 50 t/s, it is slower than DeepSeek V4 Pro 0424 (83.8 t/s) and GPT-5.2 (73.3 t/s).

Artificial Analysis Indices
Maximum
Intelligence30.1
Rank 124 of 744 measured models
Coding63.0
Rank 74 of 298 measured models
Agentic31.8
Rank 63 of 189 measured models
Reasoning levels
Each point is a reasoning level of this model: more reasoning often raises the score but costs more. Select a point to open that level.
Intelligence Indexmost efficient ↗Price / 1M tokens
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting33.5
Rank 102 of 218 measured models
Strategy & Ops30.3
Rank 115 of 218 measured models
Legal37.2
Rank 90 of 218 measured models
Healthcare & Medical31.2
Rank 52 of 82 measured models
Engineering31.4
Rank 107 of 217 measured models
Economics42.4
Rank 98 of 225 measured models
Standard Benchmarks
GPQA Diamond87.5%
Rank 101 of 619 measured models
HLE33.6%
Rank 132 of 664 measured models
SciCode50.1%
Rank 108 of 231 measured models
IFBench56.6%
Rank 146 of 450 measured models
AA-LCR80.0%
Rank 77 of 575 measured models
TerminalBench Hard53.0%
Rank 17 of 432 measured models
Terminal-Bench 2.171.2%
Rank 72 of 244 measured models
Terminal-Bench v43.0%
Rank 105 of 218 measured models
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy