← Models

Claude Opus 4.7

Anthropic · Claude 4

Anthropic's April 2026 premium Claude model, now superseded by Opus 4.8 but still useful for pinned deployments.

Part of Claude Opus family · Other versions: Claude Opus 4.8 , Claude Opus 4.6
Type
multimodal
Context
1M tokens
Max Output
128K tokens
Status
legacy
Input
$5/1M tok
Output
$25/1M tok
API Access
Yes
License
proprietary
reasoning coding agentic vision tool-use long-context
Released April 2026 · Updated May 29, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot originally verified on May 1, 2026 and rechecked for successor status on May 29, 2026.

Claude Opus 4.7 was Anthropic’s April 2026 premium model, announced on April 16, 2026 as a direct upgrade over Opus 4.6. It is no longer the newest Opus release after Claude Opus 4.8 shipped on May 28, 2026. This page remains useful for pinned deployments, benchmark comparisons, and systems that have not migrated to the newer claude-opus-4-8 API ID.

This entry is the version-specific page for Opus 4.7. For new prompts and use cases, prefer linking to the claude-opus family overview so references stay stable as new Opus snapshots ship.

Capabilities

Anthropic’s launch materials and API docs highlight a specific capability profile:

  • Strong long-horizon reasoning and agentic coding, including reductions in dropped constraints across long traces.
  • High-resolution image support up to 2576px / 3.75MP, with simpler 1:1 coordinate mapping for screenshot, document, and computer-use workflows.
  • Better file-system-based memory behavior, helpful for agents that maintain notes or scratchpads across turns.
  • More literal instruction following, calibrated response length, and a more direct tone than Opus 4.6.
  • Adaptive thinking is the only thinking-on mode; extended thinking budgets and most sampling parameters were removed in 4.7.

In practice, prompts that included scaffolding to force step-checks, restate constraints, or counteract verbosity are good candidates for re-baselining on Opus 4.7.

Technical Details

Anthropic’s official model docs and pricing pages list:

  • 1M token context window at standard API pricing, with no long-context premium.
  • 128K max output tokens.
  • Multimodal input (text and images), with high-resolution image support.
  • Adaptive thinking, the new xhigh effort level, and a beta task_budgets mode for advisory token caps across full agentic loops.
  • A new tokenizer that may produce roughly 1x to 1.35x as many tokens for the same text compared to Opus 4.6, so real-world cost can rise even though per-token rates are unchanged. Anthropic recommends adding headroom to max_tokens and compaction triggers when migrating.

API model ID: claude-opus-4-7. Aliasing rules still favor pinning to a specific version ID for strict reproducibility in production.

Pricing & Access

Base API pricing (per 1M tokens):

  • Input: $5
  • Output: $25
  • 5-minute cache writes: $6.25
  • 1-hour cache writes: $10
  • Cache hits and refreshes: $0.50
  • Batch API input: $2.50
  • Batch API output: $12.50

Long-context requests through 1M input tokens use the same per-token rate. Anthropic’s data residency option for US-only inference applies a 1.1x multiplier on all token categories. As of the May 2026 Opus 4.8 refresh, Opus 4.7 also has fast-mode pricing, but teams starting fresh should verify whether they need the older 4.7 snapshot before paying for premium speed.

Access options:

  • Anthropic Claude API as claude-opus-4-7 for pinned or migration-sensitive deployments
  • AWS Bedrock
  • Google Cloud Vertex AI
  • Microsoft Foundry
  • Claude consumer and workspace products, plus GitHub Copilot

Best Use Cases

Choose Opus 4.7 when the task justifies premium spend:

  • Multi-file or multi-repo coding work with strict correctness requirements.
  • Long-running agents that must keep plan integrity across many tool calls and recover gracefully from partial failures.
  • Vision-heavy document, diagram, and computer-use workflows that benefit from high-resolution image handling.
  • Knowledge work involving slide editing, .docx redlining, or chart and figure analysis where pixel-level fidelity matters.

For lighter or higher-volume work, Sonnet 4.6 remains the better cost-performance default. For frozen-snapshot reproducibility against older Opus behavior, the legacy Opus 4.6 entry is still the right reference.

Comparisons

  • Claude Opus 4.8 (Anthropic): Current successor with better agentic coding behavior, mid-conversation system messages, lower prompt-cache minimums, and cheaper fast mode.
  • Claude Opus 4.6 (Anthropic): Direct predecessor at the same headline price, retained mostly for reproducibility against older 4.6-era behavior.
  • Claude Sonnet 4.6 (Anthropic): Cheaper general-purpose default, with a similar 1M-context profile and lighter premium tax.
  • GPT-5.5 (OpenAI): Comparable frontier route on the OpenAI side, with stronger ChatGPT and Codex product surfaces and a broader cloud distribution story.