System Card← leaderboard
Anthropic

Claude Haiku

#16of 28 · general

Haiku 5.5

1Mcontext
$0.10 / $0.50per 1M in / out
2026-10-07released
2026-06-01knowledge cutoff
50100Coding 48 · #10 of 28Agents 30 · #23 of 29Cyber 31 · #8 of 24Engineering 46 · #11 of 26Math 52 · #9 of 26Health 62 · #6 of 24Legal 27 · #21 of 27Professional 35 · #17 of 28Writing: no readingInstruct 19 · #19 of 25Hallucination 27 · #19 of 26Multimodal 30 · #15 of 24
Coding48#10
Agents30#23
Cyber31#8
Engineering46#11
Math52#9
Health62#6
Legal27#21
Professional35#17
Writing—
Instruct19#19
Hallucination27#19
Multimodal30#15

History

AA Intelligence Index1020304050Oct 2025Jan 2026Apr 2026Jul 2026Oct 2026Claude Haiku 4.5 · $1/M in · 2025-10-15 · still served · AA Intelligence Index 16.9 AA Intelligence Index (0–100)Claude Haiku 5.5 · $0.1/M in · 2026-10-07 · still served · AA Intelligence Index 43.4 AA Intelligence Index (0–100)4.5
  • Anthropic
  • older → newer
$ / M input tokens2 of 176 priced models · price is a third-party claim, not a vendor quote

Effort

Haiku 5.5
% of best score7080901000.2×0.5×1×× the best effort’s cost · log scaleLowMedHighXhighMax

Measured

≈ imputed · amber dot = lab-self-reported among the sources
GPQA Diamond89.7%#17Epoch AI
HLE37.3%#22Artificial AnalysisEpoch AI
LiveBench72.1%#16LiveBench
SciCode55.0%#16Artificial AnalysisEpoch AI
CritPt22.6%#11Artificial AnalysisEpoch AI
APEX Agents56.1%#10Epoch AI
GDPval55.9%#13Artificial Analysis
AA Index43.4#15Artificial Analysis
Vals Index54.3#13Vals.ai
Epoch Capabilities Index159.3#8Epoch AI

Neighbors

GLM Flash→ weaker mathGLM→ follows instructions better · pricierDeepSeek Flash→ stronger agentMiMo Pro→ pricierQwen Flash→ follows instructions better · weaker coding