Anthropic Releases Claude Opus 4.7 With Coding Gains

Anthropic released Claude Opus 4.7 on April 16, calling it the most intelligent model in its Opus tier. The new release became available on Amazon Bedrock starting April 17 and shipped with measurable gains on coding, vision and long-horizon agentic tasks.

Coding and Vision Benchmarks

Opus 4.7 posted a 13% uplift over its predecessor on the SWE-Bench Pro coding benchmark, according to Anthropic and AWS. Visual acuity improved by roughly 3x, a meaningful jump for screen-reading agents and document workflows. Anthropic said agentic performance on long, multi-step tasks also improved.

Pricing Holds Steady

Pricing stayed at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.6. The continuity is significant for enterprise customers running production workloads, as it keeps unit economics stable while delivering the upgrade.

Bedrock and Context Window

The model is available in Amazon Bedrock from April 17. Separately, the Sonnet 4.6 1 million-token context window remains in beta, giving developers two distinct paths for handling very long inputs depending on whether they prioritize raw intelligence or context length.