AnthropicReleased Oct 7, 2026
Proprietary

haiku-5.5

Strongest in Mathematical Reasoning (#28 of 180) and Abstract Reasoning (#41 of 209); weakest in Reliability (#76 of 181). All seven dimensions are measured.

Composite IQ ?
122
#42 of 180 ranked models · ahead of 77%
Anthropic's #10 model, behind fable-5.2 (144)
Effective cost
$1.32/1M
$0.10 in · $0.50 out
Speed
234tok/s
360s typical response
Context
1Mtokens
1,000,000 tokens in
Evidence
7of 7 dims
24 of 39 benchmarks

IQ profile

Seven dimensions, averaged equally into Composite IQ
Compare shape with
100 avg human130160● Abstract106#41● Math131#28● Academic128#57● Coding134#36● Computer Use114#75● Reliability114#76● Knowledge Work125#37
Measured95% range
#1 in dimensionTop-10 median

IQ dimensions

Ranks are among models scored in each dimension. Ranges are model-based estimates, not calibrated confidence intervals.

DimensionIQPosition · 70–160RankEvidencevs Top 10
106measured
#41 / 209
4/5
-26
131measured
#28 / 180
3/5
-16
128measured
#57 / 211
3/6
-7
134measured
#36 / 181
3/6
-16
114measured
#75 / 180
1/5
-30.5
114measured
#76 / 181
5/6
-20.5
125measured
#37 / 180
5/6
-7
haiku-5.595% range (model-based, uncalibrated)Top-10 median#1 in dimension100 = average human

Benchmark results

24 of 39 Composite IQ benchmarks have a published result for haiku-5.5. Benchmarks without a result stay listed so gaps are visible.

Abstract Reasoning IQ 106 · measuredBest: #32 of 71 on ARC-AGI-24/5
BenchmarkScorePercentileRankLeader · source
ARC-AGI-2
Full source leaderboard reconciliation · Oct 23
61.3%
#32 / 71
Leader gpt-6-astra · 95.0%
Full source leaderboard reconciliation · Oct 23
ARC-AGI-1
Full source leaderboard reconciliation · Oct 23
89.0%
#34 / 71
Leader fable-5 · 98.5%
Full source leaderboard reconciliation · Oct 23
Mystery Game Puzzles
Full source leaderboard reconciliation vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 23
30.0%
#27 / 55
Leader gpt-6-astra · 84.0%
Full source leaderboard reconciliation vEpoch 100-position leaderboard (task implementation versions recorded per run) · Oct 23
ARC-AGI-3
Full source leaderboard reconciliation · Oct 23
0.5%
#17 / 26
Leader fable-5.2 · 81.2%
Full source leaderboard reconciliation · Oct 23
Not yet reported
Chess Puzzles
Mathematical Reasoning IQ 131 · measuredBest: #12 of 49 on ProofBench 1.13/5
BenchmarkScorePercentileRankLeader · source
ProofBench 1.1
Full source leaderboard reconciliation v1.1 · Oct 27
89.0%
#12 / 49
Leader fable-5.1 · 100%
Full source leaderboard reconciliation v1.1 · Oct 27
FrontierMath Tier 1-3
Full source leaderboard reconciliation · Oct 23
75.1%
#22 / 87
Leader fable-5.2 · 95.1%
Full source leaderboard reconciliation · Oct 23
FrontierMath Tier 4
Full source leaderboard reconciliation · Oct 23
46.3%
#23 / 75
Leader fable-5.2 · 100%
Full source leaderboard reconciliation · Oct 23
Not yet reported
MathArenaAIME
Academic Reasoning IQ 128 · measuredBest: #27 of 167 on Humanity's Last Exam3/6
BenchmarkScorePercentileRankLeader · source
Humanity's Last Exam
Artificial Analysis v4.3.2 · Oct 27
44.4%
#27 / 167
Leader opus-5.5 · 61.4%
Artificial Analysis v4.3.2 · Oct 27
CritPt
Artificial Analysis v4.3.2 · Oct 27
18.9%
#30 / 159
Leader Muse Wonder Full · 32.9%
Artificial Analysis v4.3.2 · Oct 27
SciCode
Artificial Analysis v4.3.2 · Oct 27
55.0%
#35 / 167
Leader opus-5.5 · 66.9%
Artificial Analysis v4.3.2 · Oct 27
Not yet reported
GPQA DiamondMMLU-ProMMMU-Pro
Programmatic Reasoning IQ 134 · measuredBest: #11 of 63 on ProgramBench3/6
BenchmarkScorePercentileRankLeader · source
ProgramBench
Full source leaderboard reconciliation · Oct 23
29.5%
#11 / 63
Leader opus-5.5 · 65.0%
Full source leaderboard reconciliation · Oct 23
Terminal-Bench 4.0
Artificial Analysis v5.0.1 · Nov 13
32.8%
#26 / 109
Leader Kimi K3.1 Pro · 67.2%
Artificial Analysis v5.0.1 · Nov 13
IOI V2
Full source leaderboard reconciliation v2 · Oct 27
47.3%
#29 / 43
Leader Gemini 4 Argon · 100%
Full source leaderboard reconciliation v2 · Oct 27
Not yet reported
LiveCodeBenchSWE-rebenchMirrorCode
Computer Use IQ 114 · measuredBest: #36 of 72 on Arena.ai Agent Arena1/5
BenchmarkScorePercentileRankLeader · source
Arena.ai Agent Arena
Arena.ai Agent Arena · Nov 13
0.5
#36 / 72
Leader fable-5.2 · 30.4
Arena.ai Agent Arena · Nov 13
Not yet reported
BrowseCompOSWorld-VerifiedMCP AtlasAgents' Last Exam
Reliability IQ 114 · measuredBest: #26 of 154 on AA Long Context Reasoning v1.15/6
BenchmarkScorePercentileRankLeader · source
AA Long Context Reasoning v1.1
Artificial Analysis v4.3.2 · Oct 27
82.7%
#26 / 154
Leader kimi-k3 · 88.7%
Artificial Analysis v4.3.2 · Oct 27
AA Omniscience
Artificial Analysis v4.3.2 · Oct 27
10.7
#43 / 158
Leader opus-5.5 · 46.4
Artificial Analysis v4.3.2 · Oct 27
BullshitBench v2
Full source leaderboard reconciliation · Oct 23
57.0%
#29 / 88
Leader opus-4.8 · 94.0%
Full source leaderboard reconciliation · Oct 23
FACTS Grounding
Full source leaderboard reconciliation · Nov 13
57.0%
#33 / 45
Leader gpt-6.1-terra · 84.2%
Full source leaderboard reconciliation · Nov 13
SimpleQA Verified
Full source leaderboard reconciliation · Oct 23
23.8%
#73 / 83
Leader fable-5.2 · 94.0%
Full source leaderboard reconciliation · Oct 23
Not yet reported
IFBench
Knowledge Work IQ 125 · measuredBest: #13 of 116 on AA-Briefcase v1.1 Elo5/6
BenchmarkScorePercentileRankLeader · source
AA-Briefcase v1.1 Elo
Artificial Analysis v5.0.1 · Nov 13
1,573
#13 / 116
Leader sonnet-5.5 · 1,816
Artificial Analysis v5.0.1 · Nov 13
GDPval-AA v2.1 normalized score
Artificial Analysis v4.3.2 · Oct 23
56.0%
#16 / 124
Leader opus-5.5 · 68.3%
Artificial Analysis v4.3.2 · Oct 23
AA-AnalystAgent
Full source leaderboard reconciliation vpass^5-2026-09-29 · Oct 27
53.8%
#6 / 35
Leader gemini-3.7-flash · 60.0%
Full source leaderboard reconciliation vpass^5-2026-09-29 · Oct 27
GDP.pdf
Artificial Analysis v5.0.1 · Nov 13
20.8%
#28 / 111
Leader Muse Wonder Full · 32.6%
Artificial Analysis v5.0.1 · Nov 13
AutomationBench-AA
Artificial Analysis v5.0.1 · Nov 13
35.4%
#54 / 112
Leader Gemini 4 Carbon · 77.8%
Artificial Analysis v5.0.1 · Nov 13
Not yet reported
EnterpriseOps-Gym-AA

Release & availability

Facts from the publisher's announcement and documentation.

Released
Developer
Anthropic
Access
Available through the Claude API as claude-haiku-5-5 and on Amazon Web Services, Google Cloud and Microsoft Azure.
Weights
ProprietaryHosted access only
Inputs
textimage
Context
1,000,000 tokens
API price
$0.10 in · $0.50 outper 1M tokens · full pricing

What Anthropic says

  • Adaptive thinking with adjustable effort for high-volume and latency-sensitive work.
  • Supports a 1M-token context window and 128K output tokens.

Limitations

  • Higher pricing applies to prompts over 100,000 tokens; listed token prices use the lower prompt-length tier.
  • Biology and cybersecurity safeguards restrict some requests.
  • Effort settings are configurations of one model, not separate commercial models.

Publisher statements, not AI IQ measurements.

About this score

How much of haiku-5.5's IQ is measured, and how the rest is estimated.

24/ 39benchmark results
7/ 7dimensions measured
0estimated
4/5Abstract
3/5Math
3/6Academic
3/6Coding
1/5Computer Use
5/6Reliability
5/6Knowledge Work
Bar height = share of benchmarks with resultsEstimatedNo score
Measured results are shown as published

Every score in the benchmark tables comes from its source unchanged. Composite IQ averages all seven dimensions equally.

How gaps are handled

Every dimension has at least one published result. Missing benchmarks can also draw on its predecessor, Haiku 4.5; a predecessor is an estimation input, not evidence that this model has the predecessor's measured results.

A missing result is not a zero

Benchmark versions and evaluation harnesses can differ between sources, so small gaps between models may not be meaningful.

Read the scoring methodology →

Compare haiku-5.5

Side-by-side pages for the models it is most often weighed against.

haiku-5.5Comparison model
Browse all comparisons →

Frequently asked questions

Quick answers about haiku-5.5, drawn from the same data as the rest of this page.

How AI IQ scores models →Get this data from the API →
What is the IQ of haiku-5.5?
haiku-5.5 has an estimated IQ of 122, which ranks #42 out of 180 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
What is haiku-5.5 best at?
Among the dimensions with published benchmark results, haiku-5.5 ranks highest in Mathematical Reasoning (IQ 131, #28 of 180) and Abstract Reasoning (IQ 106, #41 of 209). Its lowest-ranked measured dimension is Reliability (IQ 114, #76 of 181).
What is the EQ of haiku-5.5?
haiku-5.5 doesn't have an EQ score yet. EQ (emotional reasoning) is calculated from EQ-Bench 4 and AttuneBench, and neither has published a result for haiku-5.5. AI IQ doesn't estimate EQ from other benchmarks, so the score appears once one of those results is published.
Which benchmarks has haiku-5.5 been tested on?
haiku-5.5 has published results on 24 of the 39 benchmarks behind AI IQ's Composite IQ, including AA-Briefcase v1.1 Elo, GDPval-AA v2.1 normalized score, Humanity's Last Exam, AA Long Context Reasoning v1.1 and AA-AnalystAgent.
How much does haiku-5.5 cost?
haiku-5.5 costs $0.10 per million input tokens and $0.50 per million output tokens. Adjusted for how many tokens it actually consumes on real tasks, its effective cost works out to $1.32 per million input-plus-output tokens.
Is haiku-5.5 open source?
No — haiku-5.5 is a proprietary model. Its weights are not publicly available, so it is used through hosted APIs and products.
Who developed haiku-5.5?
haiku-5.5 was developed by Anthropic, based in the United States.
When was haiku-5.5 released?
haiku-5.5 was publicly released on October 7, 2026.
How large is the context window of haiku-5.5?
haiku-5.5 supports a context window of 1,000,000 tokens — the amount of text it can consider in a single request.
How fast is haiku-5.5?
In independent measurements, haiku-5.5 generates about 234 tokens per second, with a typical response time of about 360 seconds.