Anthropic just shipped Claude Opus 4.7, and it's a meaningful step up. Released April 16th, this model brings significant improvements to software engineering—the kind that actually matter for developers shipping real code.

What's Better

Coding: Opus 4.7 lands 13% better on Anthropic's 93-task coding benchmark and handles the multi-step, long-running tasks that usually trip up AI. Users report being able to hand off their hardest work with confidence. It catches its own logical mistakes during planning, which sounds small until you're debugging why an AI assistant missed an edge case.

Vision: 3x higher image resolution at the same API price. It's more tasteful completing professional tasks—generating interfaces, decks, documentation that actually looks good.

Honesty: Claims 92% honesty rate with fewer hallucinations and less sycophancy. The model thinks more deeply, brings opinionated perspective instead of just agreeing with you. That's a shift.

Cost and Availability

Same pricing as Opus 4.6: $5 per million input tokens, $25 per million output tokens. Available now across Claude.ai, the API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry.

The Safety Angle

Anthropic's also shipping safeguards on Opus 4.7 around cybersecurity use—automatic detection and blocking of high-risk requests. They're testing these on Opus before rolling them to Claude Mythos (their frontier model). Security professionals who want to use it for legitimate pentesting can join their new Cyber Verification Program.

Why It Matters

This is the model Anthropic's betting on for developers. The coding improvements are real—users like Rakuten report 3x better performance on production tasks. For software engineers, Opus 4.7 is the go-to on Claude's public tier right now.

Source: Anthropic - Claude Opus 4.7