Claude Opus 4.7 (April 16, 2026): /ultrareview, Auto Mode for Max, xhigh Effort Level, and a Better SWE Score

Anthropic released Claude Opus 4.7 on April 16, 2026 at the same $5 / $25 per MTok price as Opus 4.6. The headline isn't a price cut — it's a coding-quality bump plus three workflow changes inside Claude Code: a new /ultrareview slash command, auto mode extended to Max users, and a new xhigh effort level that becomes the default across all plans.

TL;DR

• Released: April 16, 2026. API model ID: claude-opus-4-7. Pricing: $5 / $25 per MTok (unchanged from 4.6).
• Coding: notable improvement on Opus 4.6 in advanced software engineering; ~13% gain on one 93-task internal benchmark (cited in the announcement), with four tasks no prior Opus or Sonnet 4.6 could solve.
• Vision: images up to 2,576 px on the long edge (~3.75 MP), more than 3× prior capacity.
• Memory: improved file system-based memory for multi-session work.
• Claude Code adds: /ultrareview slash command, auto mode for Max users, xhigh as the new default effort level.
• Available on: Claude.ai, Anthropic API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry.

Quick Answer

If you're already on Opus 4.6, the upgrade math is simple: same price, better software engineering, better long-multimodal handling, better instruction following. The Claude Code surface changes are the part most worth your attention — /ultrareview plus the v2.1.120 non-interactive claude ultrareview subcommand means cloud code review can now run in CI on top of a sharper model. xhigh becoming the default effort level means baseline output quality on Claude Code is going up even if you change nothing.

What's New in Opus 4.7

Software-engineering gains

Anthropic's framing is that Opus 4.7 has a "notable improvement on Opus 4.6 in advanced software engineering." The headline number cited in the launch is from a partner benchmark — ~13% gain on a 93-task internal coding suite, including four tasks no prior Opus or Sonnet 4.6 could solve. SWE-bench-style numbers from the official model card and benchmark partners will fill in the broader picture, but the direction is clear: 4.7 is meaningfully stronger on long-horizon coding work.

Vision: 2,576 px long edge (~3.75 MP)

Vision capacity jumps from prior 800-ish-px maxes to 2,576 pixels on the long edge — about 3.75 megapixels — more than 3× the prior limit. For developer workflows, that's the difference between giving Claude a screenshot of an entire architecture diagram or design system and giving it a thumbnail.

Better instruction following and file-system memory

Two quieter but high-leverage changes:

• Substantially better instruction following — tighter adherence to constraints, especially in multi-step plans.
• Improved file system-based memory for multi-session work — better at maintaining context across long-running projects when memory is rooted in real files rather than a transcript.

Pricing

Pricing is unchanged from Opus 4.6: $5 per million input tokens / $25 per million output tokens. Opus 4.7 is a same-price upgrade, which is the friendliest possible release shape for teams already running Opus 4.6 in production.

Claude Code Adds /ultrareview, Auto Mode for Max, and xhigh Default

The model launch lands with three Claude Code workflow additions:

/ultrareview slash command

/ultrareview is a dedicated cloud code-review session — the interactive counterpart to the non-interactive claude ultrareview CI subcommand. Run it inside Claude Code to spawn a focused review pass on the current branch or a PR target. With Opus 4.7 underneath, the review side now has a sharper model than the day-to-day claude session — a standard "use a stronger model for review than for the first draft" pattern.

Auto mode extended to Max users

Auto mode lets Claude make routine decisions on its own — choosing tools, accepting low-risk changes, advancing through plans without prompting at every step — and was previously gated to a narrower audience. With Opus 4.7, Claude Max users get auto mode in Claude Code. If you've been bouncing between explicit-prompt mode and full agent mode, auto sits in the middle.

xhigh is the new default effort level

xhigh ("extra high") slots between high and max and becomes the default effort level across all plans with Opus 4.7. Practically:

• You get higher baseline reasoning depth on Claude Code without changing your config.
• max is still available when you explicitly want it for a hard problem.
• Skills can branch on the active effort via ${CLAUDE_EFFORT} (added in v2.1.120) — useful if you have skills that should behave differently at xhigh vs low.

Availability

How Opus 4.7 Compares to Opus 4.6

For a side-by-side that includes other frontier models, see our Opus 4.5 vs GPT-5.1 Codex Max vs Gemini 3 Pro comparison and Opus 4.6 vs Opus 4.5.

Migration Notes

• Code changes: drop-in. Switch the model ID to claude-opus-4-7 in your Anthropic SDK calls; pricing math stays the same as 4.6.
• Claude Code users: run claude update to pick up /ultrareview, auto mode availability changes, and the xhigh default. Combine with the v2.1.120–v2.1.123 release for the full late-April surface.
• Effort budgeting: if you have cost-sensitive workloads, note that xhigh as the new default may use more output tokens than high. Pin effort: high (or low) where cost matters more than ceiling quality.

Sources

• Anthropic news — Claude Opus 4.7: anthropic.com/news/claude-opus-4-7
• Claude Code changelog (for /ultrareview, auto mode, xhigh): code.claude.com/docs/en/changelog
• Claude Code v2.1.120–v2.1.123 update: /blog/claude-code-2-1-120-april-2026-update

For broader model context, see our beginner's guide to top AI models, Claude vs GPT for coding, and how to use Claude Opus 4.5.