Claude Opus 4.7

Learning Objectives

• Understand Claude Opus 4.7's capabilities and what changed from Opus 4.6
• Compare Opus 4.7 with Sonnet 4.6, Haiku 4.5, and competing frontier models
• Evaluate the new xhigh effort level, task budgets, and vision improvements for practical use

What Is Claude Opus 4.7?

Claude Opus 4.7 is Anthropic's most capable generally available model, released on April 16, 2026. It replaces Opus 4.6 as the flagship — delivering a step-change improvement in agentic coding, significantly upgraded vision, and new developer controls for managing autonomous agent work.

Opus 4.7 is part of the Claude model family alongside Sonnet 4.6 (the balanced workhorse) and Haiku 4.5 (speed and efficiency). The model becomes the default for Enterprise and API users on April 23, 2026.

What Changed from Opus 4.6

Major Coding Performance Gains

Opus 4.7 delivers Anthropic's largest coding improvement between model generations:

On a 93-task internal coding benchmark, Opus 4.7 lifted resolution by 13% over Opus 4.6, including four tasks neither Opus 4.6 nor Sonnet 4.6 could solve. Code quality is noticeably improved — the model produces cleaner code, avoids unnecessary wrapper functions and fallback scaffolding, and fixes its own mistakes as it works.

3.3x Higher Resolution Vision

Opus 4.7 accepts images up to 2,576 pixels on the long edge (approximately 3.75 megapixels), compared to Opus 4.6's 1.15 megapixels. This is a 3.3x resolution increase that directly improves the model's ability to read charts, diagrams, screenshots, and documents.

• CharXiv-R (no tools): 82.1% (up from 68.7%) — isolating the vision capability itself
• Visual navigation (no tools): 79.5% (up from 57.7%)

New "xhigh" Effort Level

Opus 4.7 introduces xhigh, a new effort level sitting between high and max. This gives developers finer control over the tradeoff between reasoning depth and response latency. For coding and agentic use cases, Anthropic recommends starting at high or xhigh. In Claude Code, the default effort level has been raised to xhigh for all plans.

Task Budgets (Public Beta)

A new platform feature paired with Opus 4.7: task budgets let developers cap and prioritize token spend across long-running autonomous jobs. Combined with xhigh effort, you can tell the model "think hard on this, but don't burn more than N tokens finishing it" — essential for cost control in production agentic workflows.

New Tokenizer

Opus 4.7 ships with a new tokenizer that can produce up to 35% more tokens for the same input text compared to previous Claude models. This contributes to improved performance but means the same prompt may cost more tokens than it did with Opus 4.6. Factor this into cost planning when migrating.

The Claude Model Family

Choosing between models:

• Start with Sonnet for nearly all tasks — it handles most professional work well at lower cost
• Escalate to Opus 4.7 when Sonnet's output is materially insufficient — complex architecture decisions, multi-file refactoring, long-horizon agentic tasks, or tasks requiring high-resolution vision
• Use Haiku for high-volume production workloads where per-request cost dominates (chatbots, classification, customer service)

Pricing

API Pricing

• Input: $5 per million tokens
• Output: $25 per million tokens
• Prompt caching: Up to 90% savings on cached input tokens
• Batch processing: 50% savings on eligible requests
• No multiplier for long contexts — same per-token rate regardless of context length
• Tokenizer note: The new tokenizer may produce up to 35% more tokens for the same text, effectively increasing per-request cost even though per-token pricing is unchanged

Claude.ai Subscription Tiers

Claude Opus 4.7 vs. Competing Frontier Models

Opus 4.7 leads on both SWE-bench Verified (87.6%) and SWE-bench Pro (64.3%) among generally available models. Its vision upgrade (3.75MP) closes the gap with Gemini's traditionally strong multimodal capabilities. GPT-5.5 retains the largest third-party ecosystem and native image generation.

Products Built on Opus 4.7

Several Anthropic products are powered by Opus 4.7:

• Claude Code — Anthropic's terminal-based coding agent. The new xhigh effort level is the default in Claude Code for all plans.
• Claude Design — AI-powered visual creation tool (research preview, April 2026) for prototypes, decks, one-pagers, and UI mockups. Uses Opus 4.7's 3.75MP vision for reading brand references and design files.
• Claude Cowork — Desktop agent for autonomous file and document management.
• Claude Managed Agents — Cloud-hosted agents at scale through Anthropic's infrastructure.

Each product brings Opus 4.7's capabilities to a different surface — terminal (Code), visual design (Design), desktop (Cowork), and cloud deployment (Managed Agents).

Cybersecurity Approach

Anthropic experimented with differential reduction of Opus 4.7's cyber capabilities during training — making the model less useful for offensive cyber operations while preserving defensive security utility. Security professionals interested in using the model for legitimate cybersecurity purposes can apply through Anthropic's formal verification program.

Strengths

• Leading SWE-bench scores — 87.6% Verified, 64.3% Pro, 70% CursorBench
• 3.75 megapixel vision — 3.3x higher resolution than Opus 4.6; reads charts, screenshots, and documents with significantly better accuracy
• xhigh effort level — finer control over reasoning depth vs. latency for agentic workflows
• Task budgets — cap token spend on autonomous agents for predictable costs
• 1 million token context — full-codebase and full-document analysis with no per-token price penalty
• Cleaner code output — eliminates unnecessary wrapper functions and scaffolding
• Safety-focused design — Constitutional AI training; differential cyber capability reduction

Limitations and Considerations

• Closed model — API-only; no downloadable weights or self-hosting option
• New tokenizer costs more — up to 35% more tokens for the same input text; same per-token price but higher per-request cost
• Higher cost than Sonnet — $5/$25 vs. $3/$15 per million tokens; use Sonnet as default and escalate to Opus only when needed
• Pro plan required on claude.ai — $20/month minimum for Opus access (Free plan gets Sonnet only)
• Conservative refusals — safety training occasionally produces overly cautious refusals on legitimate requests
• Smaller ecosystem than GPT — fewer third-party tutorials and integrations compared to OpenAI's ecosystem
• No native image generation — Anthropic does not currently offer image generation capabilities

Key Takeaways

• Claude Opus 4.7 is Anthropic's most capable generally available model — leading SWE-bench Verified (87.6%), SWE-bench Pro (64.3%), and CursorBench (70%)
• Vision improved 3.3x (3.75MP) with dramatically better chart and document reading; the new xhigh effort level and task budgets give developers precise control over agentic workflows
• The new tokenizer improves performance but uses up to 35% more tokens for the same text — factor this into migration cost planning
• Becomes the default API model on April 23, 2026; Opus 4.6 remains available but is no longer the flagship