Claude Opus 4.5

// Overview

Claude Opus 4.5 is a large language model from Anthropic, released on November 24, 2025 as part of the Claude Opus line. At launch Anthropic positioned it as its most capable Opus-tier model and described it as the best model in the world for coding, agents, and computer use. It is available through the Claude apps, the Claude API, and the three major cloud platforms (Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry).

Claude Opus 4.5 accepts text and image input, ships with a 200K-token context window, and can generate up to 64K output tokens in a single synchronous response. It supports extended thinking and introduced the effort parameter, which lets developers dial how much the model reasons before responding — Anthropic reported it using 48–76% fewer output tokens than its predecessors for comparable results.

The headline change versus the prior Opus generation is price: Claude Opus 4.5 launched at $5 per million input tokens and $25 per million output tokens, roughly a two-thirds cut from Claude Opus 4.1's $15/$75. The model has since been superseded by Claude Opus 4.6, 4.7, and 4.8, so it is now a legacy model, but the API ID claude-opus-4-5-20251101 remains available.

// Benchmarks

1. SWE-bench Verified80.9%
2. Terminal-Bench59.3%
3. Aider Polyglot89.4%
4. GPQA Diamond87%
5. MMMLU (multilingual)90.8%
6. ARC-AGI-237.6%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

Roughly a two-thirds cut from Claude Opus 4.1 ($15 / $75). Cache writes and the Batch API (50% discount) are priced separately.

Pricing source ↗

// Strengths

• State-of-the-art real-world software engineering at launch — 80.9% on SWE-bench Verified, leading frontier models
• Strong agentic and computer-use performance, including 59.3% on Terminal-Bench
• The effort parameter for tuning reasoning depth against token cost (48–76% fewer output tokens for comparable results)
• Extended thinking for hard reasoning, coding, and math problems
• Large two-thirds price cut versus Opus 4.1 while raising the capability ceiling
• Robustness to prompt-injection attacks relative to other frontier models

// Best for

• Reach for it when you need a capable model for multi-step agentic coding and bug-fixing across real repositories.
• Use it for computer-use and terminal/agent workflows that chain many tool calls.
• Apply it to deep research, spreadsheet, and slide work that benefits from extended thinking.
• Pick it when you want Opus-class capability at a materially lower price than Opus 4.1.

// How to access

// Claude Opus — every version

The full lineage of the Claude Opus line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.