Claude Opus 4.6

Claude Opus 4.6 is Anthropic's most intelligent model, improving on its predecessor's coding skills with more careful planning, longer agentic task sustenance, more reliable operation in larger codebases, and better code review and debugging skills. First Opus-class model with 1M token context window (beta), 128K output tokens, and adaptive thinking.

AIME 2025

99.8%

i
Tau2 Telecom

99.3%

i
Graphwalks parents >128k

95.4%

i
MRCR v2 (8-needle)

93.0%

i
Tau2 Retail

91.9%

i
DeepSearchQA

91.3%

i
GPQA

91.3%

i
MMMLU

91.1%

i
BrowseComp

84.0%

i
SWE-Bench Verified

80.8%

i
FigQA

78.3%

i
SWE-bench Multilingual

77.8%

i
CharXiv-R

77.4%

i
MMMU-Pro

77.3%

i
CyberGym

73.8%

i
OSWorld

72.7%

i
ARC-AGI v2

68.8%

i
Terminal-Bench 2.0

65.4%

i
MCP Atlas

62.7%

i
Graphwalks BFS >128k

61.5%

i
Finance Agent

60.7%

i
Humanity's Last Exam

53.1%

i
OpenRCA

34.9%

i
Vending-Bench 2

8,017.59

i
GDPval-AA

1,606

i

Pricing, uptime, and speed via OpenRouter — updated Jul 17, 2026, 04:19 AM.

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
Amazon Bedrock	available	$5.00/Mtok cache $0.50/Mtok	$25.00/Mtok	1.0M tokens context 128K tokens max output	100.0% 5m 100.0%	2,014 ms p50 TTFT 35 tok/s p50	$0.01/web search
Anthropic	available	$5.00/Mtok cache $0.50/Mtok	$25.00/Mtok	1.0M tokens context 128K tokens max output	—	2,893 ms p50 TTFT 29 tok/s p50	$0.01/web search
Azure	available	$5.00/Mtok cache $0.50/Mtok	$25.00/Mtok	1.0M tokens context 128K tokens max output	—	—	$0.01/web search
Google	available	$5.50/Mtok cache $0.55/Mtok	$27.50/Mtok	1.0M tokens context 128K tokens max output	—	2,608 ms p50 TTFT 51 tok/s p50	$0.01/web search