Inference
EN
Anthropic

Claude Haiku 5.5

Claude Haiku 5.5 is a reasoning model from Anthropic. 50 benchmarks count toward its index, in 7 categories.

  • Closed weights
  • Reasoning
  • Input: text · image · file
  • Released 2026-10
RankedEnough results for an index.
IndexAll categories, out of 100.66.3±2.8Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.159/sMedian 67/s
First tokenSeconds to the first output token.1.6 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.10batch $0.05Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.50batch $0.25 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1MMedian 262K
EloLMArena rating and rank.1399±6 #159Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
anthropic/claude-haiku-5.5
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.01
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
6 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Amazon BedrockCloud$0.10$0.501M
AnthropicMaker$0.10$0.501M
DigitalOceanCloud$0.10$0.501M
Google Vertex AICloud$0.10$0.501M
Microsoft AzureCloud$0.10$0.501M
OpenRouterRouter$0.10$0.501M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.14 counted68.629 / 218
CodingCode writing and repair.13 counted71.018 / 209
ReasoningLogic problems and puzzles.11 counted67.834 / 227
MultimodalTasks with images and text.1 counted51.9—
KnowledgeFacts and expert knowledge.5 counted54.599 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted37.2—
MathMath problems.5 counted68.323 / 168

Results

50 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Vibe Code Bench v1.1Coding90.4% ±1.74 / 123
LiveBench DesignAgentic83.1%5 / 52
Artificial Analysis Long Context ReasoningReasoning82.7%23 / 217
OTIS Mock AIME 2024-2025Math98.9% ±0.522 / 200
GDPval-AA normalizedAgentic56.0%15 / 133
Artificial Analysis Humanity's Last ExamKnowledge44.4%28 / 222
MirrorCodeCodingAbove fitFar above this model's index.78.9% ±8.05 / 39
Critical Physics TasksReasoning18.9%33 / 214
LiveBench Long AgenticAgentic57.6%8 / 51
Chess PuzzlesReasoning55.0% ±5.016 / 90
LiveBench Coding 2.0Coding91.4%10 / 52
Artificial Analysis AnalystAgentAgentic53.8%5 / 26
GPQA diamond · Epoch AIKnowledge89.6% ±2.044 / 217
ProofBench v1.1Math89.0% ±3.112 / 54
Artificial Analysis SciCodeCoding55.0%29 / 114
FrontierMath-Tiers-1-3-v2-PrivateMath75.1% ±2.623 / 90
Code MigrationCoding44.6% ±4.220 / 77
Humanity's Last Exam without toolsKnowledge45.9%14 / 53
Medical Long Context Reasoning (MLCR-AA)Reasoning35.0%6 / 22
LiveBench ResearchAgentic81.9%15 / 52
ARC-AGI-2 (semi-private)Reasoning61.3%38 / 131
Humanity's Last Exam with toolsAgentic57.4%11 / 36
ARC-AGI-1 (semi-private)Reasoning89.0%39 / 126
LiveBench Agentic CodingAgentic57.2%22 / 66
FrontierMath-Tier-4-v2-PrivateMath46.3% ±7.924 / 72
MazeBench: visual spatial reasoning in a 3D open worldAgentic1.0%16 / 48
Terminal-Bench 4.0Agentic36.4% ±2.321 / 61
LiveBench Browser UseAgentic80.0%17 / 49
Mystery Game PuzzlesReasoning30.0% ±4.630 / 80
cursorBench40Coding48.4%8 / 21
EBR-benchReasoning43.8% ±1.812 / 31
Furniture AssemblyReasoning47.5% ±6.415 / 37
FrontierCode 2.0Coding27.7%11 / 27
FrontierCode 1.1 MainCoding46.4%7 / 17
Artificial Analysis Omniscience AccuracyKnowledge34.7%68 / 164
Vibe Code Bench 1-100Coding20.4% ±3.811 / 26
Agent Arena steerabilityAgentic2.424 / 54
FrontierSWE v2Coding43.8%10 / 22
LiveBench CodingCoding77.9%33 / 66
Agent Arena task outcomeAgentic1.628 / 54
LiveBench ReasoningReasoning83.5%45 / 80
Multimodal Knowledge Agent BenchmarkAgentic41.0%11 / 19
ARC-AGI-3 (semi-private)Reasoning0.5%16 / 26
Hermes Index: agentic capability and cost per task inside Hermes AgentAgentic37.0%10 / 16
IOICoding47.3% ±10.729 / 45
Artificial Analysis Terminal-Bench v4.0Agentic32.8%15 / 22
LiveBench MathematicsMath86.4%47 / 66
Artificial Analysis GDP.pdfAgentic20.8%16 / 22
Bug Hunt BenchCoding21.5%19 / 25
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal56.0%31 / 39
OSWorld 2.1 offline subset (82 tasks)Agentic72.4%8 / 10
SimpleQA VerifiedKnowledge23.8% ±1.375 / 91
LiveBench LanguageKnowledge62.6%63 / 66
LiveBench Instruction FollowingInstruction54.7%64 / 66
LiveBench Data AnalysisReasoning52.4%79 / 80
Artificial Analysis AutomationBenchAgentic35.4%23 / 23
MedScribeHealthcare89.5% ±1.96 / 109
SAGEEducation89.8% ±1.43 / 29
ProgramBenchCoding2.5% ±1.18 / 65
CyberBench v1.1Other75.5% ±5.26 / 48
MedCodeHealthcare51.5% ±2.215 / 106
ARC-AGI-2 (public eval)Reasoning64.4%31 / 122
Legal Research BenchLegal43.8% ±3.425 / 90
Tax Agent BenchFinance31.9% ±3.427 / 82
Finance Agent (v2)Finance54.1% ±2.128 / 79
Terminal-Bench-ScienceScience10.0% ±3.616 / 44
MysteryMechanismScience35.1% ±3.211 / 30
ARC-AGI-1 (public eval)Reasoning90.1%45 / 116
BioMysteryBenchScience70.4% ±1.313 / 30
SRE BenchOther5.7% ±1.416 / 36
EMBFinance59.0% ±4.237 / 75
Public Benefits Bench v1.1Public benefits63.2% ±1.325 / 50
HLAB-Vals-1.1Legal4.2% ±1.530 / 35
Artificial Analysis Harvey LAB-AA v1.0Agentic89.9%—
PerceptionBench with toolsMultimodal78.6%—
Hermes Bench: 150 agent tasks graded on the files and state the agent leaves behindAgentic71.0%—
Artificial Analysis HealthBench Professional raw scoreKnowledge64.8%—
Artificial Analysis HealthBench raw scoreKnowledge61.2%—
Artificial Analysis HealthBench length-adjusted scoreKnowledge58.5%—
Artificial Analysis HealthBench Professional length-adjusted scoreKnowledge58.4%—
Agents' Last Exam, Linux subsetAgentic57.0%—
Chartography without toolsMultimodal46.4%—
FrontierCode 2.0, Artificial Analysis evaluationCoding34.9%—
Artificial Analysis AutomationBench, strict pass@1Agentic13.9%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYLiveBenchNo licence statedVals AINo licence statedARC Prize FoundationTerms restrict reuseLMArenaCC BY 4.0models.devMIT

History

5 releases
Index66.9 → 66.3
23 Oct 202612 Nov 2026
Output / 1M$0.50 → $0.50
23 Oct 202612 Nov 2026

Questions

How much does Claude Haiku 5.5 cost?
$0.10 per 1M input tokens, $0.50 per 1M output tokens.
How good is Claude Haiku 5.5?
Its index is 66.3 out of 100. That is place 37 of 283 ranked models.
What is Claude Haiku 5.5 best at?
Coding: 71.0 out of 100, place 18 of 209.
How fast is Claude Haiku 5.5?
159 output tokens per second, 1.6 s to the first token.
What is the context window of Claude Haiku 5.5?
1M tokens.
Is Claude Haiku 5.5 open source?
No. Its weights are not published.
When was Claude Haiku 5.5 released?
October 2026.

Badge

Claude Haiku 5.5 on Inference360: place 37 of 283, index 66.3
HTML
<a href="https://inference360.ai/models/claude-haiku-5-5"><img src="https://inference360.ai/badges/claude-haiku-5-5.svg" alt="Claude Haiku 5.5 on Inference360: place 37 of 283, index 66.3" height="20"></a>
Markdown
[![Claude Haiku 5.5 on Inference360: place 37 of 283, index 66.3](https://inference360.ai/badges/claude-haiku-5-5.svg)](https://inference360.ai/models/claude-haiku-5-5)

Compare with

Inkling 2 Small66.4GLM-5.365.8Qwen3.8 Max65.6Claude Sonnet 5.577.2Claude Fable 5.285.6

Lower price

dots3-note Preview63.4 · FreeLing 3.1 Flash67.2 · $0.135/1MMiMo-V2.6-Flash64.3 · $0.28/1M

More from Anthropic

All models
Claude Sonnet 5.577.2Claude Opus 5.578.6Claude Fable 5.285.6Claude Mythos 5.2UnrankedClaude Fable 5.176.2