Skip to content
Model Pareto

Claude Haiku 5.5 vs Claude Haiku 4.5

Claude Haiku 5.5 scores higher than Claude Haiku 4.5 in all 3 shared categories. Claude Haiku 4.5 costs 10× as much as Claude Haiku 5.5 ($2.00 vs $0.20 per million tokens, blended 3:1 input:output).
Claude Haiku 5.5Claude Haiku 4.5
Focus

Coding: quality vs price

Compared models are ringed; the other 117 ranked here are greyed.

Lower price is better. Pareto frontier: Nex-N2.5-Mini, Nex-N2.5-Pro, MiMo-V2.6-Flash, Ternary Bonsai 2 27B, MiMo-V2.6-Pro, Gemini 4 Argon, Claude Opus 5.5. 34 models are not plotted: GLM-5.3 Flash NVFP4, DeepSeek V4.1 Flash NVFP4, GLM-5.3 NVFP4, DeepSeek-V4-Pro-0813-nvfp4-DSpark, Xing4.0-29B-A4B, Ternary Bonsai 27B, Atria Dawn Preview, Maple-Preview, Bonsai 27B, Qwen3.8 27B NVFP4, Intern-S2-397B, Phi-4-reasoning-plus, Solar Open2 250B, Laguna S 2.1, Motif 3, Nex-N2-Pro, K-EXAONE 2.0 0803, K2 Horizon 375B A23B, A.X-K2, K2 Horizon MoVA 36B A4B, Intern-S2-Preview (35B-A3B), North Mini Code, G9v3-39A5B, Falcon-H1-34B-Instruct, Qwen3.8 Flash-Next NVFP4, Devstral 2, K2 Horizon 7B, K2-Horizon-32B, Devstral Small 2, MiniCPM5-2B, Ling 3.0 Tiny, K2 Horizon 3.7B, K2 Horizon 0.9B, LFM2.5-2.6B.

34 models with no price data — shown in the strip at the left edge

GLM-5.3 Flash NVFP4, DeepSeek V4.1 Flash NVFP4, GLM-5.3 NVFP4, DeepSeek-V4-Pro-0813-nvfp4-DSpark, Xing4.0-29B-A4B, Ternary Bonsai 27B, Atria Dawn Preview, Maple-Preview, Bonsai 27B, Qwen3.8 27B NVFP4, Intern-S2-397B, Phi-4-reasoning-plus, Solar Open2 250B, Laguna S 2.1, Motif 3, Nex-N2-Pro, K-EXAONE 2.0 0803, K2 Horizon 375B A23B, A.X-K2, K2 Horizon MoVA 36B A4B, Intern-S2-Preview (35B-A3B), North Mini Code, G9v3-39A5B, Falcon-H1-34B-Instruct, Qwen3.8 Flash-Next NVFP4, Devstral 2, K2 Horizon 7B, K2-Horizon-32B, Devstral Small 2, MiniCPM5-2B, Ling 3.0 Tiny, K2 Horizon 3.7B, K2 Horizon 0.9B, LFM2.5-2.6B

  • Best-value frontier (nothing is both cheaper and better)
  • Evidencestrong → weak
  • Estimated from other categories
  • Claude Haiku 5.5
    Quality
    66.2
    Rank
    #19/119
    Price
    $0.20/M tok
    Speed
    —
    In$0.10▾Out$0.50▾/M tok

    vs Claude Haiku 4.5: +34.2 quality · 0.1× price

  • Claude Haiku 4.5
    Quality
    32.0
    Rank
    #79/119
    Price
    $2.00/M tok
    Speed
    109tok/s
    In$1.00Out$5.00/M tok

    vs Claude Haiku 5.5: −34.2 quality · 10× price

Overview

Metric
Claude Haiku 5.5
Claude Haiku 4.5
Lab
Anthropic
Anthropic
Released
Oct 7, 2026
Oct 15, 2025
Weights
Proprietary
Proprietary
Input price
$/M tok
$0.10Best
$1.00
Output price
$/M tok
$0.50Best
$5.00
Blended price
$/M tok · 3:1 in:out
$0.20Best
$2.00
Output speed
tok/s
—
109
Context
tokens
1MBest
200K

Quality by category

0–100 within each category (100 = best tracked model). Rank is among all models ranked there. Click a row to focus it.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
Coding
66.2±8.0Best
#19/119Verified
32.1±5.3
#79/119Mixed
Writing & chat
Not ranked
47.6±3.8
#72/156Mixed
Vision
78.7±9.7Best
#7/91Lab-reportedbest value
36.2±4.1
#66/91Mixed
Hard reasoning
81.3±5.1Best
#17/200Verifiedbest value
40.8±3.4
#103/200Verified
Agents
Not ranked
22.3±3.5
#117/178Mixed

Coding benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
SWE-bench Verified
day 0
~#2/14~86.8%Estimated
  1. #1Claude Opus 596.0%
  2. ~#2Claude Haiku 5.5Estimated86.8%
  3. #2Claude Sonnet 585.2%
estimated from related benchmarks
#10/1373.3%Lab-reported
  1. #1Claude Opus 596.0%
  2. #9Xing4.0-29B-A4B75.0%
  3. #10Claude Haiku 4.573.3%
  4. #11Nemotron 3 Ultra70.7%
anthropic.com · 2025-10-15
SWE-bench Pro (Public)
day 0
~#6/20~68.1%Estimated
  1. #1Claude Opus 5.589.9%
  2. #5Intern-S2-397B68.5%
  3. ~#6Claude Haiku 5.5Estimated68.1%
  4. #6Hy4 preview65.7%
estimated from related benchmarks
~#19/20~50.8%Estimated
  1. #1Claude Opus 5.589.9%
  2. #18Muse Glimmer51.2%
  3. ~#19Claude Haiku 4.5Estimated50.8%
  4. #19Nex-N2.5-Mini43.8%
estimated from related benchmarks
DeepSWE v1.1
day 0
~#22/35~63.2%Estimated
  1. #1Gemini 4 Argon78.8%
  2. #21GLM-5.3 Flash63.4%
  3. ~#22Claude Haiku 5.5Estimated63.2%
  4. #22Claude Opus 562.5%
estimated from related benchmarks
~#31/35~47.8%Estimated
  1. #1Gemini 4 Argon78.8%
  2. #30Qwen3.8 Max (0902)51.0%
  3. ~#31Claude Haiku 4.5Estimated47.8%
  4. #31Qwen3.8 27B42.2%
estimated from related benchmarks
Terminal-Bench 4.0
day 0
#15/8832.8%BestVerified
  1. #1Claude Sonnet 5.563.6%
  2. #13Step 5 Preview33.3%
  3. #15Claude Haiku 5.532.8%
  4. #15GLM-5.3 Flash32.8%
artificialanalysis.ai · 2026-10-07
#53/880.0%Verified
  1. #1Claude Sonnet 5.563.6%
  2. #53Celeris-10.0%
  3. #53Claude Haiku 4.50.0%
  4. #53Devstral 20.0%
artificialanalysis.ai · 2026-09-24
LiveCodeBench
third-party
~#7/25~84.4%Estimated
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #6gpt-oss-120b87.8%
  3. ~#7Claude Haiku 5.5Estimated84.4%
  4. #7Ternary Bonsai 27B82.8%
estimated from related benchmarks
#13/2461.5%Verified
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #12Mistral Small 464.0%
  3. #13Claude Haiku 4.561.5%
  4. #14GLM-5.3 NVFP453.5%
artificialanalysis.ai · 2026-09-24
SciCode
third-party
#21/10555.0%BestVerified
  1. #1Claude Opus 5.566.9%
  2. #20GPT-6.1 Sol55.7%
  3. #21Claude Haiku 5.555.0%
  4. #21GPT-5.6 Terra55.0%
artificialanalysis.ai · 2026-10-07
#60/10542.2%Verified
  1. #1Claude Opus 5.566.9%
  2. #59Ling-3.0-flash-Fin42.4%
  3. #60Claude Haiku 4.542.2%
  4. #60Motif 342.2%
artificialanalysis.ai · 2026-09-24
AA Coding Agent Index
third-party
~#12/18~50.9Estimated
  1. #1Claude Opus 5.566.0
  2. #11Kimi K351.9
  3. ~#12Claude Haiku 5.5Estimated50.9
  4. #12Mistral Large 449.8
estimated from related benchmarks
~last/18below measured range (<41.1)Estimated
  1. #1Claude Opus 5.566.0
  2. #17GPT-6 Luna41.1
  3. ~#18Claude Haiku 4.5Estimated<41.1
estimated from related benchmarks
Benchmarks in other categories (4)

Writing & chat benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
IFBench
day 0
—
#63/13354.3%Verified
  1. #1MiniMax-M382.9%
  2. #62Falcon-H1R-7B54.4%
  3. #63Claude Haiku 4.554.3%
  4. #64Jamba Reasoning 3B52.4%
artificialanalysis.ai · 2026-09-24
MMMLU
day 0
—
#3/483.0%Lab-reported
  1. #1Gemini 3.1 Pro92.6%
  2. #2Gemma 4 31B88.4%
  3. #3Claude Haiku 4.583.0%
  4. #4gpt-oss-120b81.3%
anthropic.com · 2025-10-15
LMArena Text, Non-English (Elo)
third-party
—
#41/561395Verified
  1. #1Gemini 4 ArgonFew votes1512
  2. #39Qwen3.5 122B A10B1397
  3. #41Claude Haiku 4.51395
  4. #42GPT-5.4 nano1387
lmarena.ai · 2026-09-24
LMArena Text (Elo)
third-party
—
#40/561415Verified
  1. #1Gemini 4 Argon1525
  2. #39Qwen3.5 122B A10B1417
  3. #40Claude Haiku 4.51415
  4. #41Mistral Large 31413
lmarena.ai · 2026-09-24
EQ-Bench Creative Writing v3 (Elo)
third-party
—
~#19/26~1580Estimated
  1. #1GPT-6 Astra2173
  2. #18Inkling1611
  3. ~#19Claude Haiku 4.5Estimated1580
  4. #19Gemini 3.5 Flash-Lite1559
estimated from related benchmarks

Vision benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
MMMU
day 0
~#2/3~71.3%Estimated
  1. #1Claude Haiku 4.573.2%
  2. ~#2Claude Haiku 5.5Estimated71.3%
  3. #2Llama 4 Maverick69.4%
estimated from related benchmarks
#1/273.2%Lab-reported
  1. #1Claude Haiku 4.573.2%
  2. #2Llama 4 Maverick69.4%
anthropic.com · 2025-10-15
MMMU-Pro
day 0
~#18/86~79.1%Estimated
  1. #1Claude Opus 5.587.7%
  2. #17Apodex 1.179.2%
  3. ~#18Claude Haiku 5.5Estimated79.1%
  4. #18Gemini 3.5 Flash-Lite79.0%
estimated from related benchmarks
#62/8558.6%Verified
  1. #1Claude Opus 5.587.7%
  2. #61Bonsai 27B60.5%
  3. #62Claude Haiku 4.558.6%
  4. #63Mistral Small 456.8%
artificialanalysis.ai · 2026-09-24
CharXiv Reasoning
day 0
~#3/9~84.6%Estimated
  1. #1Gemini 3.8 Flash86.2%
  2. #2Kimi K384.8%
  3. ~#3Claude Haiku 5.5Estimated84.6%
  4. #3Qwen3.8 Flash-Next84.6%
estimated from related benchmarks
~last/9below measured range (<74.5%)Estimated
  1. #1Gemini 3.8 Flash86.2%
  2. #8Gemini 3.5 Flash-Lite74.5%
  3. ~#9Claude Haiku 4.5Estimated<74.5%
estimated from related benchmarks
OmniDocBench v1.5
third-party
~#4/8~89.2%Estimated
  1. #1MiniMax-M391.6%
  2. #2Qwen3.8 27B91.1%
  3. ~#4Claude Haiku 5.5Estimated89.2%
  4. #4Gemini 3.1 Pro85.3%
estimated from related benchmarks
#5/779.6%Verified
  1. #1MiniMax-M391.6%
  2. #4Gemini 3.1 Pro85.3%
  3. #5Claude Haiku 4.579.6%
  4. #6Mistral Small 476.4%
benchmarking.nanonets.com · 2026-09-24
GDP.pdf (AA)
third-party
~#11/44~21.3%Estimated
  1. #1GPT-6.1 Sol32.0%
  2. #10Claude Opus 521.6%
  3. ~#11Claude Haiku 5.5Estimated21.3%
  4. #11GPT-5.521.2%
estimated from related benchmarks
#34/433.8%Verified
  1. #1GPT-6.1 Sol32.0%
  2. #33Qwen3.6 35B A3B6.6%
  3. #34Claude Haiku 4.53.8%
  4. #35Qwen3.5 122B A10B3.4%
artificialanalysis.ai · 2026-09-24
Chartography
third-party
#4/2346.4%Lab-reported
  1. #1GPT-6 Astra71.0%
  2. #3GPT-6 Sol53.6%
  3. #4Claude Haiku 5.546.4%
  4. #5Claude Fable 5.146.2%
anthropic.com · 2026-10-07
~last/24below measured range (<9.0%)Estimated
  1. #1GPT-6 Astra71.0%
  2. #23Mistral Large 39.0%
  3. ~#24Claude Haiku 4.5Estimated<9.0%
estimated from related benchmarks

Hard reasoning benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
GPQA Diamond
day 0
~#3/184~95.0%Estimated
  1. #1GPT-6 Astra96.1%
  2. #2Gemini 3.8 Flash95.3%
  3. ~#3Claude Haiku 5.5Estimated95.0%
  4. #3Gemini 3.1 Pro94.1%
estimated from related benchmarks
#112/18367.2%Verified
  1. #1GPT-6 Astra96.1%
  2. #111Mistral Large 368.0%
  3. #112Claude Haiku 4.567.2%
  4. #113Llama 4 Maverick67.1%
artificialanalysis.ai · 2026-09-24
Humanity's Last Exam
day 0
#16/18444.4%BestVerified
  1. #1Claude Opus 5.561.4%
  2. #15GPT-5.545.8%
  3. #16Claude Haiku 5.544.4%
  4. #17Grok 4.644.1%
artificialanalysis.ai · 2026-10-07
#104/18410.4%Verified
  1. #1Claude Opus 5.561.4%
  2. #103K2-V210.5%
  3. #104Claude Haiku 4.510.4%
  4. #104Solar Open 100B10.4%
artificialanalysis.ai · 2026-09-24
AIME (latest)
day 0
~#2/32~97.9%Estimated
  1. #1GLM-5.3 NVFP499.5%
  2. ~#2Claude Haiku 5.5Estimated97.9%
  3. #2Inkling97.1%
estimated from related benchmarks
#18/3183.7%Verified
  1. #1GLM-5.3 NVFP499.5%
  2. #17Mistral Small 484.0%
  3. #18Claude Haiku 4.583.7%
  4. #19Magistral Small 1.280.3%
artificialanalysis.ai · 2026-09-24
AA Intelligence Index
third-party
#17/8543.4BestVerified
  1. #1Claude Opus 5.557.6
  2. #16Kimi K343.6
  3. #17Claude Haiku 5.543.4
  4. #18GPT-5.6 Terra42.1
artificialanalysis.ai · 2026-10-07
#55/8516.9Verified
  1. #1Claude Opus 5.557.6
  2. #54Muse Glimmer17.5
  3. #55Claude Haiku 4.516.9
  4. #56Ring-2.6-1T16.6
artificialanalysis.ai · 2026-09-24

Agents benchmarks

Rank among models with a published score, and the leaderboard around each model. ~ italic = no published score, estimated from related benchmarks.

Metric
Claude Haiku 5.5
Claude Haiku 4.5
τ²-bench (Telecom)
day 0
—
#47/11754.7%Verified
  1. #1Step 3.7 Flash98.5%
  2. #46Gemma 4 31B59.9%
  3. #47Claude Haiku 4.554.7%
  4. #48Nemotron Cascade 2 30B A3B53.2%
artificialanalysis.ai · 2026-09-24
OSWorld-Verified
day 0
—
#11/1150.7%Lab-reported
  1. #1Kimi K384.8%
  2. #10Muse Glimmer65.9%
  3. #11Claude Haiku 4.550.7%
anthropic.com · 2025-10-15
OSWorld 2.0
day 0
—
~last/5below measured range (<47.5%)Estimated
  1. #1GLM-5.3 Flash59.1%
  2. #4GPT-5.547.5%
  3. ~#5Claude Haiku 4.5Estimated<47.5%
estimated from related benchmarks
BrowseComp
day 0
—
~#15/18~56.7%Estimated
  1. #1Atria Dawn Preview92.5%
  2. #14Inkling77.1%
  3. ~#15Claude Haiku 4.5Estimated56.7%
  4. #15Mistral Medium 3.548.6%
estimated from related benchmarks
GDPval-AA (Elo)
third-party
—
#55/87719Verified
  1. #1Claude Opus 5.51866
  2. #54KAT Coder Pro V2720
  3. #55Claude Haiku 4.5719
  4. #56MiniCPM5-2B694
artificialanalysis.ai · 2026-09-24
AutomationBench-AA
third-party
—
#54/813.2%Verified
  1. #1Claude Opus 5.569.5%
  2. #53Nemotron 3 Super 120B A12B3.8%
  3. #54Claude Haiku 4.53.2%
  4. #55Devstral 23.1%
artificialanalysis.ai · 2026-09-24
τ-Bench Banking (AA)
third-party
—
#63/899.3%Verified
  1. #1Muse Spark 1.350.5%
  2. #62Nemotron 3 Super 120B A12B10.3%
  3. #63Claude Haiku 4.59.3%
  4. #63Qwen3.6 35B A3B9.3%
artificialanalysis.ai · 2026-09-24