Claude Opus 4.7: What Developers Need to Know

Anthropic released Claude Opus 4.7 on April 16, 2026 — a model that doesn't reinvent Claude but sharpens it exactly where developers need it most: sustained coding, vision, and instruction following.

What's actually new

13% coding benchmark lift. Not a marketing number — this translates to meaningfully harder tasks resolved autonomously. Early reports say engineers are handing Opus 4.7 their hardest multi-file refactors and getting clean results back.

3.75 megapixel vision. Previous Claude models topped out around 1MP. Opus 4.7 processes images up to 2,576 pixels on the long edge — enough to read dense screenshots, extract data from complex diagrams, and work with detailed UI mockups.

New xhigh effort level. Sitting between high and max, this gives developers finer control over the trade-off between reasoning depth and latency. For complex tasks where high isn't quite enough but max is overkill, xhigh hits the sweet spot.

Task budgets (public beta). You can now cap token spend on autonomous agents, preventing runaway bills on long-running jobs. Set a budget, and Claude prioritizes work within it.

New tokenizer. Opus 4.7 uses an updated tokenizer that processes text differently — expect 1.0–1.35x more tokens for the same input, depending on content type. Factor this into cost estimates.

The model lineup as of today

Model	Price (MTok)	Context	Best for
Opus 4.7	$5 in / $25 out	1M tokens	Complex reasoning, agentic coding, production
Sonnet 4.6	$3 in / $15 out	1M tokens	Balanced speed + intelligence, daily driver
Haiku 4.5	$1 in / $5 out	200K tokens	Fast, cheap, high-volume automation

All three support text + image input, multilingual output, and tool use. Opus 4.7's knowledge cutoff is January 2026 — the most current of any Claude model.

What this means for Claude Code

Opus 4.7 is already the default in Claude Code. The xhigh effort level is available via /effort or the --effort flag. The new /ultrareview command uses Opus 4.7 for dedicated code review sessions.

If you're on a Claude Max subscription, the extended auto mode now lets Claude make more decisions on your behalf with fewer confirmation prompts — useful for long-running tasks where you want to step away.

Migration notes

If you're calling the API directly:

• Model ID: claude-opus-4-7
• Pricing unchanged from Opus 4.6
• The new tokenizer means your existing token count estimates may be slightly off
• Important: Opus 4.7 follows instructions more literally than previous models. Prompts written for earlier versions may produce unexpected results. Test before deploying.

The Mythos question

Anthropic acknowledged that Opus 4.7 trails their internal Claude Mythos Preview model, which remains invitation-only as part of Project Glasswing for defensive cybersecurity. Opus 4.7 intentionally has reduced cyber capabilities compared to Mythos. For most developers, this doesn't matter — Opus 4.7 is the strongest commercially available model for coding and general reasoning.

Should you upgrade?

If you're on Opus 4.6: yes, immediately. Same price, better at everything, especially long-running agentic tasks.

If you're on Sonnet 4.6: depends on your use case. Sonnet remains the better daily driver for cost-conscious production workloads. Opus 4.7 is worth it when you need the absolute best reasoning — complex refactors, architectural decisions, or tasks that require sustained attention across large codebases.

If you're teaching yourself AI: this is a great time to start. The Prompt Engineering and Building with the Claude API courses on this site cover everything from first principles to production deployment, updated for Opus 4.7.