Back

Anthropic

Claude 4.5 Sonnet

ReasoningClosed weights
Intelligence
20.7
Rank 236 of 744 measured models
Coding
52.1
Rank 115 of 298 measured models
Output speed
44.1 t/s
Rank 342 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
$0.557
Rank 110 of 193 measured models

Claude 4.5 Sonnet (Reasoning) scores 20.7 on the Artificial Analysis Intelligence Index, ranking 236th of 744 measured models. The closest models on this index are GPT-5.4 nano (20.7), GPT-5 Mini (batch) (20.6), and K2 Horizon 7B (20.6). At $6 per million tokens (3:1 input/output blended price), it is more expensive than GPT-5.4 nano ($0.463). With an output speed of 44.1 t/s, it is slower than GPT-5.4 nano (166 t/s).

Artificial Analysis Indices
Reasoning
Intelligence20.7
Rank 236 of 744 measured models
Coding52.1
Rank 115 of 298 measured models
Math88.0
Rank 37 of 270 measured models
Agentic15.8
Rank 108 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 & Accounting21.9
Rank 149 of 218 measured models
Strategy & Ops19.3
Rank 146 of 218 measured models
Legal25.6
Rank 145 of 218 measured models
Engineering22.8
Rank 149 of 217 measured models
Economics30.5
Rank 152 of 225 measured models
Standard Benchmarks
MMLU Pro87.5%
Rank 8 of 349 measured models
GPQA Diamond83.4%
Rank 168 of 619 measured models
HLE17.8%
Rank 240 of 664 measured models
LiveCodeBench71.4%
Rank 62 of 343 measured models
SciCode45.7%
Rank 139 of 231 measured models
AIME 202588.0%
Rank 37 of 270 measured models
IFBench57.3%
Rank 142 of 450 measured models
AA-LCR72.3%
Rank 191 of 575 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