Claude Opus 4.7

Overview. Claude Opus 4.7 is Anthropic's frontier-tier instruct model in the Claude Opus 4 family, offered as a closed-weight API with a one-million-token context window. It accepts text and image input and returns text output, and supports adjustable reasoning-effort levels rather than exposing traditional temperature or top_p/top_k sampling controls. Developers reach for it when a task demands deep, extended reasoning over very large inputs rather than quick turnaround.

Key strengths. Artificial Analysis's intelligence index places Opus 4.7 at 55, up from 53.5 in the prior evaluation cycle, with a GPQA-Diamond score of 0.91 and an HLE (Humanity's Last Exam) score of 0.42, indicating strong performance on graduate-level and expert-level reasoning questions. Its coding index of 73.6 has held steady across repeated evaluations, reflecting consistent strength on software-engineering tasks. The model exposes multiple reasoning-effort tiers — max, xhigh, high, medium, and low — giving developers explicit control over how much extended thinking is applied to a given request.

When to use it. Use Opus 4.7 for complex, long-horizon reasoning and coding work where quality matters more than speed — deep code review, multi-step technical analysis, or research-style tasks that benefit from its high GPQA and HLE scores. The one-million-token context window makes it a strong fit when a task requires reasoning across very large documents or codebases in a single pass. Dial the reasoning-effort parameter down from the default 'high' toward 'medium' or 'low' for tasks that don't need maximum depth.

Limitations. Median time-to-first-token runs around 17 seconds with output throughput near 52 tokens per second, so this model is built for batch and asynchronous workloads rather than real-time interaction. It does not support temperature, top_p, or top_k sampling controls, and Anthropic has not disclosed a training data cutoff for this model. Anthropic has since released a newer Opus-tier model, so teams evaluating the current frontier option should be aware Opus 4.7 is no longer Anthropic's newest release in this tier.

Sovereignty note. Claude Opus 4.7 is created by Anthropic, a US company, so Inference run through this endpoint can fall under CLOUD Act jurisdiction. It is not classified as a Sovereign, EU-resident option, which is worth noting for regulated workloads in finance, healthcare, government, or legal contexts.