Noometry

Anthropic, proprietary

Claude Haiku 4.5

Claude Haiku 4.5 by Anthropic ranks 165th of 354 ranked models on the Noometry Index as of October 2026, with a score of 39.5. Its strongest category is agentic & tool use, where it ranks 52nd. API pricing starts at $1 per million input tokens and $5 per million output tokens, with a 200K-token context window.

Last verified

Noometry rank
#165 of 354
Index score
39.5
Evidence
Confirmed 53 results
Provider
Anthropic
Released
October 15, 2025
Weights
Proprietary
Reasoning
Yes
Context window
200K
Max output
64K
Input price
$1 / M
Output price
$5 / M
Blended price
$2 / M
Output speed
Not measured
Value
#151 of 219
Knowledge cutoff
February 2025
Input
text, image, pdf

Category scores

Each category score combines every public result we have in that category.

  1. Coding 44.0
  2. Agentic & Tool Use 33.6
  3. Reasoning 15.1
  4. Math 44.9
  5. Knowledge 37.7
  6. Multimodal 26.8
  7. Multilingual 49.9
  8. Instruction Following 71.4
  9. Long Context 43.6
  10. Writing & Preference 57.9
CategoryScoreRankResults
Coding44.0#786
Agentic & Tool Use33.6#525
Reasoning15.1#3208
Math44.9#784
Knowledge37.7#1536
Multimodal26.8#1181
Multilingual49.9#1291
Instruction Following71.4#1492
Long Context43.6#921
Writing & Preference57.9#1235

Strengths and weaknesses

Categories where Claude Haiku 4.5 places highest and lowest among the models ranked in each, with its score against that category's median.

Strongest categories

CategoryScorevs medianRank
Coding44.0+5.3#78 of 340, top 23%
Math44.9+8.3#78 of 327, top 24%
Long Context43.6+2.7#92 of 296, top 32%

Weakest categories

CategoryScorevs medianRank
Multimodal26.8−11.7#118 of 128, top 93%
Reasoning15.1−8.6#320 of 350, top 92%
Instruction Following71.4+0.1#149 of 305, top 49%

Closest competitors

The models ranked just above and below Claude Haiku 4.5. When scores are this close, price and speed are often the better way to choose.

ModelRankScoreBlended $/MSpeed
Nova 2 Lite#16139.7$0.85—Compare
DeepSeek-V3.2-Speciale#16239.7$0.85—Compare
Hunyuan Turbo 0110#16339.6——Compare
Claude 3.7 Sonnet#16439.5——Compare
DeepSeek-V3#16639.5$0.4173Compare
Grok 4 Fast#16739.4—577Compare
Olmo 3.1 32b Instruct#16839.4——Compare
Granite 4.2 3b#16939.4——Compare

Sponsored placements are available on pages like this one. Advertise on Noometry

Benchmark results

Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.

Coding

BenchmarkScorePositionSettingSourceDate
SWE-bench Verified (bash only)66.6%#13 of 39, top 34%highSWE-bench2026-02-17
LMArena WebDev1330#94 of 113, top 84%LMArena2026-10-08
SWE-bench Multilingual64.7%#11 of 13, top 85%SWE-bench2026-02-13
SciCode43.3%#64 of 121, top 53%Epoch AI
WeirdML45.4%#61 of 119, top 52%Epoch AI
WeirdML44.1%16KEpoch AI
LMArena Coding1453#83 of 294, top 29%LMArena2026-10-08
ALE-Bench653.48#75 of 105, top 72%32KEpoch AI

Agentic & Tool Use

BenchmarkScorePositionSettingSourceDate
Terminal-Bench35.5%#28 of 41, top 69%Epoch AI
Berkeley Function Calling Leaderboard68.7%#6 of 49, top 13%fcBerkeley Function Calling Leaderboard
DeepResearch Bench45.5%#15 of 24, top 63%lowEpoch AI
BALROG31.2%#17 of 35, top 49%Epoch AI
BALROG31.2%1KEpoch AI
ExploitBench13.7%#8 of 9, top 89%Epoch AI
Vending-Bench 2458.89#51 of 60, top 85%Epoch AI

Reasoning

BenchmarkScorePositionSettingSourceDate
ARC-AGI-21.3%Epoch AI
ARC-AGI-22.8%16KEpoch AI
ARC-AGI-21.3%1KEpoch AI
ARC-AGI-24%#61 of 83, top 74%32KEpoch AI
ARC-AGI-21.7%8KEpoch AI
NYT Connections (extended)14.3%#84 of 91, top 93%Lech Mazur benchmarks
ARC-AGI-114.3%Epoch AI
ARC-AGI-137.3%16KEpoch AI
ARC-AGI-116.8%1KEpoch AI
ARC-AGI-147.7%#58 of 83, top 70%32KEpoch AI
ARC-AGI-125.5%8KEpoch AI
CritPt0%#102 of 134, top 77%Epoch AI
Chess Puzzles8%#84 of 129, top 66%32KEpoch AI2026-07-16
LMArena Hard Prompts1420#105 of 297, top 36%LMArena2026-10-08
DTBench73.6%#90 of 151, top 60%Epoch AI
LMCA30.9%#71 of 125, top 57%Epoch AI
Epoch Capabilities Index142.41#98 of 213, top 47%Epoch AI2025-10-15
ForecastBench61.4#9 of 72, top 13%Epoch AI

Math

BenchmarkScorePositionSettingSourceDate
OTIS Mock AIME 2024-202535.8%Epoch AI2025-10-16
OTIS Mock AIME 2024-202566.7%#97 of 173, top 57%32KEpoch AI2025-10-22
Omni-MATH56.1%#12 of 57, top 22%HELM Capabilities
LMArena Math1396#130 of 285, top 46%LMArena2026-10-08
MATH Level 586.9%Epoch AI2025-10-16
MATH Level 596.4%#9 of 79, top 12%32KEpoch AI2025-10-22
FrontierMath (Feb 2025 set)4.1%Epoch AI2025-10-16
FrontierMath (Feb 2025 set)5.9%#43 of 68, top 64%32KEpoch AI2025-10-22
FrontierMath Tier 4 (v1)2.1%#39 of 55, top 71%32KEpoch AI2025-10-22

Knowledge

BenchmarkScorePositionSettingSourceDate
GPQA Diamond60.5%Epoch AI2025-10-16
GPQA Diamond71.2%#97 of 186, top 53%32KEpoch AI2025-10-22
SimpleQA Verified13.2%#70 of 77, top 91%Epoch AI2026-08-10
SimpleQA Verified12.6%32KEpoch AI2026-08-27
MMLU-Pro77.7%#25 of 58, top 44%HELM Capabilities
Vectara Hallucination Rate (lower is better)9.8%#54 of 96, top 57%Vectara Hallucination Leaderboard
GPQA (HELM)60.5%#23 of 57, top 41%HELM Capabilities
LMArena Expert1442#81 of 273, top 30%LMArena2026-10-08

Multimodal

BenchmarkScorePositionSettingSourceDate
Blueprint-Bench 20%#27 of 31, top 88%Epoch AI
LMArena Document1420#32 of 38, top 85%LMArena2026-09-13

Multilingual

BenchmarkScorePositionSettingSourceDate
LMArena Non-English1377#129 of 297, top 44%LMArena2026-10-08
LMArena Chinese1417#128 of 285, top 45%LMArena2026-10-08
LMArena French1408#114 of 223, top 52%LMArena2026-10-08
LMArena German1375#113 of 231, top 49%LMArena2026-10-08
LMArena Japanese1339#106 of 211, top 51%LMArena2026-10-08
LMArena Korean1347#106 of 213, top 50%LMArena2026-10-08
LMArena Russian1381#128 of 283, top 46%LMArena2026-10-08
LMArena Spanish1420#90 of 226, top 40%LMArena2026-10-08

Instruction Following

BenchmarkScorePositionSettingSourceDate
IFEval80.1%#43 of 57, top 76%HELM Capabilities
LMArena Instruction Following1414#80 of 298, top 27%LMArena2026-10-08

Long Context

BenchmarkScorePositionSettingSourceDate
LMArena Longer Query1427#83 of 291, top 29%LMArena2026-10-08

Writing & Preference

BenchmarkScorePositionSettingSourceDate
LMArena Text1396#122 of 297, top 42%LMArena2026-10-08
LMArena Creative Writing1372#109 of 295, top 37%LMArena2026-10-08
WildBench83.9%#15 of 57, top 27%HELM Capabilities
EQ-Bench 41064#26 of 28, top 93%EQ-Bench
LMArena Multi-Turn1409#111 of 295, top 38%LMArena2026-10-08

API pricing by provider

RouteInput $/MOutput $/MCached input $/MChecked
anthropic$1$5$0.102026-10-10
azure$1$5$0.102026-10-10
bedrock$1$5$0.102026-10-10
openrouter$1$5$0.102026-10-10
vertex$1$5$0.102026-10-10

All Anthropic API prices → Estimate your cost →

Compare Claude Haiku 4.5

Other Anthropic models

Frequently asked questions

How good is Claude Haiku 4.5?

Claude Haiku 4.5 by Anthropic ranks 165th of 354 ranked models on the Noometry Index as of October 2026, with a score of 39.5. Its strongest category is agentic & tool use, where it ranks 52nd. API pricing starts at $1 per million input tokens and $5 per million output tokens, with a 200K-token context window.

How much does Claude Haiku 4.5 cost?

Claude Haiku 4.5 costs $1 per million input tokens and $5 per million output tokens on Anthropic's own API, with cached input at $0.10.

What is Claude Haiku 4.5's context window?

Claude Haiku 4.5 accepts up to 200K tokens of input and can write up to 64K tokens in one response.

Is Claude Haiku 4.5 open source?

No. Claude Haiku 4.5 is proprietary and available only through Anthropic's API and partner platforms.

What are Claude Haiku 4.5's strengths and weaknesses?

Relative to other ranked models, Claude Haiku 4.5 places best in coding, math, long context and lowest in multimodal, reasoning, instruction following.

What is Claude Haiku 4.5 best at?

Its best category is agentic & tool use, where it ranks 52nd on Noometry.