Noometry

Anthropic, proprietary

Claude 3 Haiku

Claude 3 Haiku by Anthropic ranks 340th of 354 ranked models on the Noometry Index as of October 2026, with a score of 25.9. Its strongest category is multimodal, where it ranks 128th.

Last verified

Noometry rank
#340 of 354
Index score
25.9
Evidence
Confirmed 37 results
Provider
Anthropic
Released
March 7, 2024
Weights
Proprietary
Reasoning
Unknown
Context window
—
Max output
—
Input price
Not listed
Output price
Not listed
Blended price
Not listed
Output speed
41 tokens/s Kagi
Value
Not ranked
Knowledge cutoff
Unknown

Category scores

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

  1. Coding 26.4
  2. Reasoning 16.3
  3. Math 9.8
  4. Knowledge 17.3
  5. Multimodal 23.6
  6. Multilingual 36.0
  7. Instruction Following 61.3
  8. Long Context 36.1
  9. Writing & Preference 29.7
CategoryScoreRankResults
Coding26.4#3255
Reasoning16.3#3074
Math9.8#3193
Knowledge17.3#2853
Multimodal23.6#1281
Multilingual36.0#2431
Instruction Following61.3#2471
Long Context36.1#2371
Writing & Preference29.7#2914

Strengths and weaknesses

Categories where Claude 3 Haiku places highest and lowest among the models ranked in each, with its score against that category's median.

Strongest categories

CategoryScorevs medianRank
Long Context36.1−4.9#237 of 296, top 81%
Instruction Following61.3−10.0#247 of 305, top 81%
Multilingual36.0−11.4#243 of 297, top 82%

Weakest categories

CategoryScorevs medianRank
Multimodal23.6−14.9#128 of 128, top 100%
Math9.8−26.7#319 of 327, top 98%
Coding26.4−12.3#325 of 340, top 96%

Closest competitors

The models ranked just above and below Claude 3 Haiku. When scores are this close, price and speed are often the better way to choose.

ModelRankScoreBlended $/MSpeed
Qwen3-1.7B#33626.6——Compare
Mistral Nemo#33726.4$0.15—Compare
Ministral 3B#33826.2$0.10—Compare
DeepSeek-R1-Distill-Qwen-1.5B#33926.1——Compare
Gemma 2 9B#34125.9——Compare
Dolly 2.0-12b#34225.5——Compare
GPT-4o mini#34325.5$0.26120Compare
Llama 3-8B#34425.5——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
WeirdML9.8%#114 of 119, top 96%Epoch AI
BigCodeBench Instruct39.4%#34 of 64, top 54%BigCodeBench2024-03-07
LMArena Coding1199#241 of 294, top 82%LMArena2026-10-08
BigCodeBench Complete50.1%#34 of 66, top 52%BigCodeBench2024-03-07
CadEval12%#14 of 14, top 100%Epoch AI
HumanEval+68.9%#22 of 45, top 49%mar 2024EvalPlus
MBPP+68.8%#17 of 38, top 45%mar 2024EvalPlus

Reasoning

BenchmarkScorePositionSettingSourceDate
Kagi LLM Benchmark34.2%#89 of 99, top 90%Kagi LLM Benchmark
LMArena Hard Prompts1174#247 of 297, top 84%LMArena2026-10-08
DTBench50.1%#136 of 151, top 91%Epoch AI
LMCA8.8%#115 of 125, top 92%Epoch AI
Epoch Capabilities Index118.35#177 of 213, top 84%Epoch AI2024-03-07
ForecastBench53.2#70 of 72, top 98%Epoch AI
WinoGrande74.2%#25 of 43, top 59%Epoch AI

Math

BenchmarkScorePositionSettingSourceDate
OTIS Mock AIME 2024-20251.8%#163 of 173, top 95%Epoch AI2025-02-25
LMArena Math1188#235 of 285, top 83%LMArena2026-10-08
MATH Level 514.9%#68 of 79, top 87%Epoch AI2025-01-27

Knowledge

BenchmarkScorePositionSettingSourceDate
GPQA Diamond36.3%#156 of 186, top 84%Epoch AI2025-01-27
Confabulations (lower is better)34.2%#48 of 51, top 95%Lech Mazur benchmarks
LMArena Expert1148#236 of 273, top 87%LMArena2026-10-08
MMLU73.8%#36 of 81, top 45%Epoch AI

Multimodal

BenchmarkScorePositionSettingSourceDate
LMArena Vision950#122 of 122, top 100%LMArena2026-10-09
ScienceQA72%#3 of 6, top 50%Epoch AI

Multilingual

BenchmarkScorePositionSettingSourceDate
LMArena Non-English1178#243 of 297, top 82%LMArena2026-10-08
LMArena Chinese1155#245 of 285, top 86%LMArena2026-10-08
LMArena French1195#194 of 223, top 87%LMArena2026-10-08
LMArena German1174#198 of 231, top 86%LMArena2026-10-08
LMArena Japanese1102#186 of 211, top 89%LMArena2026-10-08
LMArena Korean1109#188 of 213, top 89%LMArena2026-10-08
LMArena Russian1204#232 of 283, top 82%LMArena2026-10-08
LMArena Spanish1166#202 of 226, top 90%LMArena2026-10-08

Instruction Following

BenchmarkScorePositionSettingSourceDate
LMArena Instruction Following1173#247 of 298, top 83%LMArena2026-10-08

Long Context

BenchmarkScorePositionSettingSourceDate
LMArena Longer Query1190#246 of 291, top 85%LMArena2026-10-08

Writing & Preference

BenchmarkScorePositionSettingSourceDate
LMArena Text1195#244 of 297, top 83%LMArena2026-10-08
LMArena Creative Writing1157#248 of 295, top 85%LMArena2026-10-08
EQ-Bench Creative Writing717#108 of 115, top 94%EQ-Bench
LMArena Multi-Turn1190#241 of 295, top 82%LMArena2026-10-08

Compare Claude 3 Haiku

Other Anthropic models

Frequently asked questions

How good is Claude 3 Haiku?

Claude 3 Haiku by Anthropic ranks 340th of 354 ranked models on the Noometry Index as of October 2026, with a score of 25.9. Its strongest category is multimodal, where it ranks 128th.

Is Claude 3 Haiku open source?

No. Claude 3 Haiku is proprietary and available only through Anthropic's API and partner platforms.

How fast is Claude 3 Haiku?

Claude 3 Haiku generated about 41 output tokens per second in the Kagi LLM Benchmark's timed runs. Speed varies by provider, load and reasoning effort.

What are Claude 3 Haiku's strengths and weaknesses?

Relative to other ranked models, Claude 3 Haiku places best in long context, instruction following, multilingual and lowest in multimodal, math, coding.

What is Claude 3 Haiku best at?

Its best category is multimodal, where it ranks 128th on Noometry.