priorsBuilder

Anthropic · released 2026-10-07

Haiku 5.5

Measured on 62 benchmarks across 5 reasoning efforts.

General frontier

44.8

#45 of 136 models · best at max

includes estimates: 44% of the mix is measured for this model

Reasoning efforts in your mix

max44.8#111 of 313
xhigh44.2#115 of 313
high41.7#130 of 313
medium37.9#163 of 313
low33.5#198 of 313

Your mix’s benchmarks

4 of 8 measured

reasoning

Humanity's Last Exam

587 results
  1. #50max of 58744.4%
  2. #60xhigh of 58742.7%
  3. #102high of 58737.3%
  4. #126medium of 58733.8%
  5. #172low of 58727.0%

Leader: Opus 5.5 · 61.4%

knowledge

AA-Omniscience

526 results
  1. #91max of 52610.7
  2. #99xhigh of 5266.0
  3. #101high of 5265.8
  4. #107medium of 5264.5
  5. #112low of 5263.3

Leader: Opus 5.5 · 46.4

reasoning

BullshitBench v2

196 results
  1. #56max of 19657.0%
  2. #69low of 19653.0%

Leader: Opus 4.8 · 95.0%

math

FrontierMath Tier 4 (v2)

70 results
  1. #24max of 7046.3%

Leader: GPT-6.1 Sol · 100.0%

All 62 benchmarks

show 58 more

agentic

AA-AnalystAgent

44 results
  1. #6max of 4453.8%

Leader: Gemini 3.7 Flash · 60.0%

agentic

AA-Briefcase v1.1

224 results
  1. #20max of 2241578
  2. #28xhigh of 2241533
  3. #46high of 2241442
  4. #58medium of 2241372
  5. #100low of 2241112

Leader: Sonnet 5.5 · 1823

agentic

AutomationBench (Zapier)

115 results
  1. #65medium of 11512.0%

Leader: Gemini 4 · 51.3%

agentic

AutomationBench-AA

219 results
  1. #101xhigh of 21936.0%
  2. #104max of 21935.4%
  3. #108high of 21933.7%
  4. #113medium of 21928.6%
  5. #126low of 21922.9%

Leader: Gemini 4 · 77.5%

agentic

BioMysteryBench

27 results
  1. #12max of 2770.4%

Leader: Sonnet 5.5 · 81.1%

agentic

Harvey's Legal Agent Benchmark

79 results
  1. #57max of 791.3%

Leader: Muse Spark 1.2 · 25.4%

agentic

Legal Research Bench

78 results
  1. #22max of 7843.3%

Leader: Muse Spark 1.3 · 55.3%

agentic

Tax Agent Bench

69 results
  1. #20max of 6931.9%

Leader: Fable 5.1 · 49.2%

agentic

Terminal-Bench 4.0 (Vals)

47 results
  1. #14max of 4735.4%

Leader: Opus 5.5 · 65.2%

agentic

Terminal-Bench Science (Vals)

40 results
  1. #16max of 4010.0%

Leader: GPT-6 Astra · 62.9%

agentic

Vending-Bench 2

67 results
  1. #45max of 67$3,602

Leader: GPT-6 Astra · $15,515

coding

Arena WebDev

122 results
  1. #29high of 1221587

Leader: Opus 5.5 · 1813

coding

Code Migration

77 results
  1. #21max of 7744.6%

Leader: Sonnet 5.5 · 69.8%

coding

FrontierCode 1.1 Extended

130 results
  1. #43max of 13058.4%
  2. #45xhigh of 13058.1%
  3. #58high of 13055.9%
  4. #64medium of 13055.0%
  5. #93low of 13048.2%

Leader: Opus 5.5 · 65.3%

coding

FrontierCode 1.1 Main

130 results
  1. #39max of 13046.4%
  2. #43xhigh of 13045.8%
  3. #62high of 13041.9%
  4. #63medium of 13041.6%
  5. #93low of 13034.8%

Leader: Opus 5.5 · 54.6%

coding

FrontierSWE v2

21 results
  1. #8max of 2143.8%

Leader: GPT-6 Astra · 65.5%

coding

IOI

44 results
  1. #29max of 4447.3%

Leader: Gemini 4 · 100.0%

coding

SciCode

226 results
  1. #52max of 22655.0%
  2. #87xhigh of 22651.7%
  3. #115low of 22649.2%
  4. #117medium of 22649.0%
  5. #118high of 22648.7%

Leader: Opus 5.5 · 66.9%

coding

Terminal-Bench Science-AA

46 results
  1. #15max of 4620.0%
  2. #16xhigh of 4618.1%
  3. #25high of 4610.0%
  4. #30medium of 467.1%
  5. #39low of 461.9%

Leader: GPT-6 Astra · 63.3%

coding

Terminal-Bench v4-AA

216 results
  1. #40max of 21632.8%
  2. #46xhigh of 21629.3%
  3. #57high of 21621.7%
  4. #67medium of 21615.2%
  5. #79low of 21612.6%

Leader: Sonnet 5.5 · 63.6%

coding

Vibe Code Bench 1-100

22 results
  1. #7max of 2220.4%

Leader: Opus 5.5 · 30.4%

coding

Vibe Code Bench v1.1

105 results
  1. #3max of 10590.4%

Leader: Sonnet 5.5 · 92.4%

composite

AA Intelligence Index v4.3.2

610 results
  1. #45max of 61043.4
  2. #55xhigh of 61041.2
  3. #78high of 61037.8
  4. #88medium of 61034.5
  5. #124low of 61029.5

Leader: Opus 5.5 · 57.6

composite

Vals Index

47 results
  1. #18max of 4754.3%

Leader: Gemini 4 · 68.9%

cost

AA Intelligence Index Cost per Task

177 results
  1. #16low of 177$0.02
  2. #22medium of 177$0.05
  3. #33high of 177$0.08
  4. #44xhigh of 177$0.12
  5. #59max of 177$0.21

Leader: GPT-6 Luna · $0.00

cost

Cost to Run AA Intelligence Index

177 results
  1. #12low of 177$34.31
  2. #19medium of 177$56.36
  3. #23high of 177$94.07
  4. #38xhigh of 177$158
  5. #54max of 177$330

Leader: GPT-6 Luna · $10.63

creative

threejseval

34 results
  1. #22high of 341434

Leader: Opus 5.5 · 2079

creative

VoxelBench (text)

55 results
  1. #21max of 551754

Leader: GPT-6 Astra · 2732

economics

Excel Modeling Benchmark

74 results
  1. #36max of 7459.0%

Leader: Fable 5.1 · 76.7%

economics

Finance Agent v2

78 results
  1. #26max of 7854.1%

Leader: Gemini 4 · 65.4%

economics

GDP.pdf

210 results
  1. #58max of 21020.8%
  2. #73xhigh of 21018.2%
  3. #82high of 21017.2%
  4. #92medium of 21015.2%
  5. #116low of 21011.2%

Leader: Muse Wonder Full (Max) · 32.6%

economics

GDPval-AA v2.1

288 results
  1. #24max of 2881620
  2. #47xhigh of 2881511
  3. #70high of 2881418
  4. #101medium of 2881274
  5. #130low of 2881122

Leader: Opus 5.5 · 1866

knowledge

AA-Omniscience Accuracy

526 results
  1. #147max of 52636.4%
  2. #150xhigh of 52634.9%
  3. #154high of 52634.8%
  4. #157medium of 52634.0%
  5. #161low of 52633.3%

Leader: Fable 5.1 · 67.2%

knowledge

AA-Omniscience Non-hallucination

526 results
  1. #76max of 52659.6%
  2. #90xhigh of 52655.6%
  3. #91high of 52655.5%
  4. #93medium of 52655.2%
  5. #95low of 52655.1%

Leader: MiniCPM5-1B (Non-reasoning) · 99.1%

knowledge

Harvey LAB-AA v1.1

27 results
  1. #14xhigh of 272.8%
  2. #16low of 271.7%
  3. #17high of 271.1%
  4. #19max of 270.6%
  5. #20medium of 270.6%

Leader: Muse Wonder Full (Max) · 13.1%

knowledge

MedCode

94 results
  1. #15max of 9451.5%

Leader: Opus 5 · 63.6%

knowledge

MedScribe

96 results
  1. #6max of 9689.5%

Leader: Opus 5.5 · 91.4%

knowledge

Public Benefits Bench v1.1

50 results
  1. #25max of 5063.2%

Leader: Opus 5 · 76.9%

long-context

AA-LCR v1.1

523 results
  1. #36max of 52382.7%
  2. #109xhigh of 52378.3%
  3. #127high of 52377.3%
  4. #128medium of 52377.3%
  5. #189low of 52372.0%

Leader: Kimi K3 · 88.7%

long-context

MLCR-AA

94 results
  1. #5max of 9463.3%

Leader: Sonnet 5.5 · 75.0%

math

FrontierMath Tiers 1–3 (v2)

115 results
  1. #18max of 11575.1%

Leader: GPT-6.1 Sol · 93.7%

math

ProofBench v1.1

52 results
  1. #11max of 5289.0%

Leader: Sonnet 5.5 · 100.0%

other

T3 Code usage share

85 results
  1. #25max of 850.0%

Leader: Opus 5.5 · 29.1%

reasoning

Blueprint-Bench 2

28 results
  1. #13max of 2833.0%

Leader: Gemini 4 · 54.4%

reasoning

CritPt

529 results
  1. #53xhigh of 52922.6%
  2. #64max of 52918.9%
  3. #65high of 52918.6%
  4. #99medium of 52912.9%
  5. #121low of 5299.1%

Leader: Muse Wonder Full (Max) · 32.9%

reasoning

Furniture Assembly

31 results
  1. #10max of 3147.5%

Leader: Opus 5.5 · 83.3%

reasoning

MysteryMechanism

27 results
  1. #10max of 2735.1%

Leader: GPT-6 Astra · 53.2%

reasoning

Roboflow Visual Reasoning

102 results
  1. #23high of 10274.8%
  2. #42low of 10268.9%

Leader: GPT-6 Astra · 91.2%

reasoning

SAGE

80 results
  1. #38max of 8045.8%

Leader: Opus 4.7 · 56.1%

reasoning

SimpleQA Verified

47 results
  1. #34max of 4723.9%

Leader: Gemini 3.1 Pro · 77.5%

reliability

Tool-call error rate

171 results
  1. #108max of 1712.2%

Leader: GPT-5.4 Pro · 0.0%

reliability

Uptime (7d)

181 results
  1. #23max of 181100.0%

Leader: GPT-5 Pro · 100.0%

safety

CyberBench v1.1

45 results
  1. #6max of 4575.5%

Leader: Wonder · 80.4%

safety

SRE Bench

32 results
  1. #14max of 325.7%

Leader: GPT-6 Astra · 56.9%

speed

Latency (Time to First Token)

351 results
  1. #246low of 3518.9s
  2. #259medium of 35113s
  3. #292high of 35125s
  4. #325xhigh of 35179s
  5. #347max of 351315s

Leader: Gemini 2.5 Flash-Lite · 0.3s

speed

Output Speed (tokens/s)

351 results
  1. #23max of 351236.4
  2. #54xhigh of 351188.1
  3. #61medium of 351178.7
  4. #65high of 351176.3
  5. #70low of 351171.0

Leader: Celeris-1 · 1522.5

speed

Provider latency

177 results
  1. #105max of 1771.5s

Leader: Qwen 2.5 7B · 0.2s

speed

Provider throughput

176 results
  1. #4max of 176145.8

Leader: Muse Spark 1.1 · 204.3