Anthropic shipped Claude Opus 4.7 today. The model ID is claude-opus-4-7, it’s live across the API, Claude Pro, Max, Team, and Enterprise, and the pricing hasn’t changed — still $5 input / $25 output per million tokens, same as Opus 4.6.
So what actually changed? Stronger coding, better vision, and improved performance on complex multi-step tasks. Anthropic describes it as “more thorough and consistent on difficult work.” That’s vague but familiar framing — it means less dropping the ball on long chains of reasoning. If you’re running Opus for deep technical work, that matters. For a full breakdown of the Claude model lineup, check our Claude AI guide.
What’s new in Claude Opus 4.7
The gains are focused. Coding is better — Anthropic calls it out explicitly, and Bolt (their app-builder integration) is reporting up to 10% improvement on longer app builds, with no regressions. Vision has improved too, though Anthropic hasn’t published numbers there. The overall positioning is that Opus 4.7 handles difficult, multi-step work more reliably than 4.6 did.
At the same time, Anthropic has cleaned house on older versions. Opus 4 and 4.1 are gone from the model selector and Claude Code. If you had workflows depending on those, it’s time to migrate.
On GitHub Copilot, 4.7 is rolling out now and replaces both Opus 4.5 and 4.6 in the model picker. Worth flagging: Opus carries a 7.5x premium multiplier on Copilot Pro+. That’s not new to 4.7, but it adds up fast if you’re using Copilot heavily. If you’re weighing Opus against Sonnet for day-to-day work, our Sonnet vs Opus comparison breaks down when the cost difference actually makes sense.
Availability and pricing
The usual places: Claude.ai (Pro, Max, Team, Enterprise), the Anthropic API, Amazon Bedrock, Vertex AI, and Microsoft Foundry. There’s also a US-only inference option at a 1.1x price multiplier — presumably lower latency routing through US infrastructure.
Pricing is flat versus 4.6: $5 per million input tokens, $25 per million output tokens. No increase for the upgrade. Opus is already expensive relative to Sonnet, so holding the line here is the right call.
The Mythos caveat
Here’s the honest part. CNBC reported that Anthropic itself acknowledges Claude Opus 4.7 trails the unreleased Claude Mythos. That’s a strange thing to ship alongside a new flagship — a public concession that your best available model is already behind something you haven’t released yet.
I don’t think that makes 4.7 a bad release. The gap between “best available now” and “best in the pipeline” has always existed. But it does set expectations. If you’re waiting for a real step-change at the top of Anthropic’s lineup, Mythos is probably that release. Opus 4.7 looks more like a solid incremental improvement — better on the things Opus was already good at, fewer rough edges on complex work. Which, honestly, is fine.
The /ultrareview command
/ultrareview is a new Claude Code command that runs a deep review of your codebase — architecture, security, performance, and maintainability — in a single pass. Pro and Max users get three free sessions included.
A standard code review checks correctness. /ultrareview is more like a senior engineer spending an afternoon with your codebase. It looks at how components interact, flags patterns that will hurt you in six months, identifies security surface area you might not be thinking about, and surfaces maintainability debt.
When to use it vs a standard review: standard review is fine for a PR or a function-level change. /ultrareview makes sense before a major refactor, after a sprint of rapid feature work, or when you’re onboarding a new engineer and want to document architecture decisions properly.
Should you upgrade?
If you’re on the API, yes. Same pricing, straight upgrade, no regressions. The model ID is claude-opus-4-7. Bolt’s data on longer app-building tasks is encouraging for anyone running it in agentic loops.
If you’re on Claude.ai, it’s rolling out now. You should see it replacing 4.6 in the model picker shortly if it isn’t there already.
If you’re holding out for Mythos, I get it. But 4.7 is the best Claude available right now, and Anthropic seems to be moving fast enough that the wait probably won’t be long.
Related: Claude Code in 2026: Pricing, Capabilities, and the Honest Verdict — Full breakdown of Claude Code costs, the Pro→Max upgrade trigger, and how it compares to Cursor and GitHub Copilot.
Related: Claude AI in 2026: Complete Guide to Models, Pricing & Use — The full Claude model lineup, pricing tiers, and when to use each model.
Anthropic released Claude Opus 4.7 today, April 16, 2026. It’s generally available right now across all Claude plans and the API — no waitlist, no limited preview. Two months after Opus 4.6 launched, this is a meaningful step up, not a maintenance release.
If you’re evaluating where it fits in your stack, here’s what actually changed.
For context on the broader model landscape, we covered the best AI models in 2026 earlier this year — Opus 4.7 changes some of that picture.