Model details
Claude Opus 4.6 is designed as a high-performance model focused on complex reasoning and sustained agentic operations. Its architecture is specifically tuned to handle large codebases, providing enhanced capabilities for code review, debugging, and self-correction. By improving its planning logic, the model can manage extended, multi-step tasks more reliably, making it well-suited for autonomous multitasking in professional environments. It excels in multidisciplinary reasoning and technical analysis, demonstrating state-of-the-art performance on benchmarks like Terminal-Bench 2.0 and Humanity’s Last Exam.
The model represents a significant evolution in the Opus lineage, building upon the foundations of its predecessor to deliver substantial gains in economically valuable knowledge work. Its development emphasizes practical utility in domains such as finance and legal research, where it has shown marked improvements in handling intricate, domain-specific tasks. With its ability to operate effectively within expansive information environments, the model is engineered to support sophisticated document creation and data synthesis, positioning it as a robust tool for users requiring high-level intelligence for complex, long-form projects.
Transparent token rates
Rates are shown per one million tokens. Combined means one million input plus one million output tokens.
Amazon Bedrock
Select your cookie preferences We use essential cookies and similar tools that are necessary to provide our site and services. We use performance cookies to collect anonymous statistics, so we can understand how customers use our site and make improvements. Essential cookies cannot be deactivated, but you can choose “C
Amazon Bedrock
Anthropic's new flagship model beat every benchmark we threw at it, eats tokens like a hungry teenager, and showed its reasoning out loud.
Amazon Bedrock
Claude Opus 4.6 has been evaluated as a frontier LLM in enterprise and security contexts, with quantitative benchmarks drawn from the Corecraft-Expanded environment within EnterpriseGym. In that environment, the Standard variant achieved a 22.10% pass rate while the Adaptive variant reached 26.00%, compared against GPT Performance analysis identifies notable failure modes including poor search strategies, incomplete tool exploration, and challenges in multi-step planning under complex workflow demands. Additional evaluation data comes from TEE-RedBench, which tested the model in adversarial red-teaming scenarios within Trusted Execut
Amazon Bedrock
The BenchLM profile reports that Claude Opus 4.6 was released in February 2026 with a 1M-token context window, non-reasoning classification, and API pricing of $5 per million input tokens and $25 per million output tokens. It assigns Opus 4.6 an overall capability score of 68/100, ranking it 27th of 230 tracked models, The page also lists Coding rank 31 out of 148 and reports throughput figures of 37 tokens per second with a 2.22-second first-token latency. BenchLM explicitly flags Opus 4.6 as superseded by Claude Opus 4.7 (Adaptive), signaling that the model is no longer the newest in its line. The data snapshot is dated September 2
Amazon Bedrock
Discover more about what's new at AWS with Claude Opus 4.6 now available in Amazon Bedrock
