Name: Claude Opus 5.5
Author: Anthropic

Claude Opus 5.5

Anthropic · 2026-09-22

Claude Opus 5.5 is Anthropic's September 22, 2026 model for long-running coding and knowledge work. Adaptive thinking is always on, with low, medium, high, xhigh, and max effort and medium as the default. Production safeguards can route flagged requests to fallback models. The published knowledge cutoff is June 2026, without day precision.

Benchmark scores

Benchmark	Score
AA-Briefcase v1.1 Elo	1822
AA-LCR v1.1	84.7
AA-Omniscience Index	46.4
Artificial Analysis Intelligence Index v4.3.2	57.6
Artificial Analysis Omniscience Accuracy	66.2
Artificial Analysis Omniscience Hallucination Rate	58.6
AutomationBench 1.0.6 (Zapier)	40.0
AutomationBench 1.0.6 HR	24.2
AutomationBench 1.0.6 Marketing	43.0
AutomationBench 1.0.6 Operations	62.0
AutomationBench 1.0.6 Sales	42.7
AutomationBench 1.0.6 Support	36.0
AutomationBench-AA 1.0.6	69.5
Capability	165.9
CritPt (avg@5)	31.7
FrontierCode 1.1 Extended - Pass Rate	70.8
FrontierCode 1.1 Extended - Score	65.3
FrontierCode 1.1 Main - Pass Rate	60.0
FrontierCode 1.1 Main - Score	54.6
GDP.pdf (Artificial Analysis, All-pass)	28.8
GDP.pdf (Artificial Analysis, Mean Criteria)	82.3
GDPval-AA v2.1	1846
Harvey LAB-AA	91.2
Humanity's Last Exam (Text-Only)	61.4
MMMU Pro	87.7
Next.js AI Agent Evaluations	97.0
SciCode 1.0.1 (Artificial Analysis)	66.9
Terminal-Bench 4.0 (Artificial Analysis)	59.6