Anthropic logo

Claude Haiku 4.5

Anthropic · Released Oct 1, 2025

ProprietarybudgetagentsCompare this model

$1

Input price / MTok

#12 of 27

$5

Output price / MTok

#13 of 27

200K

Context window

#22 of 28

98

Tokens / sec

#13 of 21

73.3

SWE-bench Verified

#8 of 12

80.7

AIME 2025

#8 of 10

73.2

MMMU

#7 of 8

50.7

OSWorld

#8 of 8

Summary

Claude Haiku 4.5 is a proprietary language model from Anthropic, released on Oct 1, 2025.

At $1 input and $5 output per million tokens, it ranks #12 of 27 for input price, placing it mid-range on input cost.

Its context window of 200K tokens ranks #22 of 28 for context depth in this dataset.

It targets budget, agents workloads.

Throughput is measured at 98 tokens per second (ranked #13 of 21 for throughput).

Fastest Claude model; best for high-volume and real-time tasks.

Benchmarks

BenchmarkScoreMax%RankSource
SWE-bench Verified73.310073.3%#8/12Anthropic (self-reported in launch annou...
AIME 202580.710080.7%#8/10Anthropic (self-reported in launch bench...
MMMU73.210073.2%#7/8DataCamp citing Anthropic official launc...
OSWorld50.710050.7%#8/8Anthropic (self-reported in launch annou...

Similar Models

Compare all 6 models

Models with the most shared benchmark coverage and closest scores to Claude Haiku 4.5.

Claude Haiku 4.5
AnthropicClaude Sonnet 4.6
OpenAIo3
GoogleGemini 2.5 Pro
OpenAIo4-mini
OpenAIGPT-5
SWE-bench Verified#8 of 12
Real-world GitHub issue resolution
AIME 2025#8 of 10
Advanced math competition problems
MMMU#7 of 8
Multimodal understanding across 30 subjects
OSWorld#8 of 8
Computer-use and GUI automation

Compare Claude Haiku 4.5 side by side

Pre-filled with 5 similar models

Open comparison

Speed source: https://artificialanalysis.ai/models/claude-4-5-haiku