Back

Anthropic

Claude 4.5 Haiku

ReasoningClosed weights
Intelligence
16.9
Rank 283 of 744 measured models
Coding
43.9
Rank 144 of 298 measured models
Output speed
85.9 t/s
Rank 207 of 403 measured models
Price/1M
$2
Rank 466 of 659 measured models
Context window
200K
Rank 618 of 1199 measured models
Cost per task
$0.277
Rank 82 of 193 measured models

Claude 4.5 Haiku (Reasoning) scores 16.9 on the Artificial Analysis Intelligence Index, ranking 283rd of 744 measured models. The closest models on this index are Ling-2.6-1T (17.0), Doubao Seed Code (16.9), and GPT-5 mini (16.8). At $2 per million tokens (3:1 input/output blended price), it is more expensive than Ling-2.6-1T ($0.85) and GPT-5 mini ($0.688). With an output speed of 85.9 t/s, it is slower than GPT-5 mini (147 t/s).

Artificial Analysis Indices
Reasoning
Intelligence16.9
Rank 283 of 744 measured models
Coding43.9
Rank 144 of 298 measured models
Math83.7
Rank 52 of 270 measured models
Agentic8.0
Rank 123 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 & Accounting14.7
Rank 171 of 218 measured models
Strategy & Ops10.1
Rank 172 of 218 measured models
Legal18.9
Rank 164 of 218 measured models
Healthcare & Medical15.7
Rank 73 of 82 measured models
Engineering19.5
Rank 162 of 217 measured models
Economics26.4
Rank 163 of 225 measured models
Standard Benchmarks
MMLU Pro76.0%
Rank 168 of 349 measured models
GPQA Diamond67.2%
Rank 348 of 619 measured models
HLE10.4%
Rank 328 of 664 measured models
LiveCodeBench61.5%
Rank 114 of 343 measured models
SciCode42.2%
Rank 160 of 231 measured models
AIME 202583.7%
Rank 52 of 270 measured models
IFBench54.3%
Rank 158 of 450 measured models
AA-LCR74.3%
Rank 164 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