Claude Opus 4 - In-Depth Overview

Anthropic · Claude

Current

Flagship

Latest in family

Model ID: claude-opus-4-20250514

Get all the details on Claude Opus 4, an AI model from Anthropic. This page covers its token limits, pricing structure, key capabilities such as multimodal_input, citations, tool_use, available API code samples, and performance strengths.

Key Metrics

Input Limit

200K tokens

Output Limit

32K tokens

Input Cost

$15.00/1M

Output Cost

$75.00/1M

Sample API Code

import anthropic

client = anthropic.Anthropic(
    api_key="YOUR_ANTHROPIC_API_KEY",
)

message = client.messages.create(
    model="claude-opus-4-20250514",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Hello, Claude!"}
    ]
)
print(message.content)

Required Libraries

anthropic

@anthropic-ai/sdk

Benchmarks

Benchmark	Score	Source	Notes
livebench global average	79.53	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench reasoning average	90.47	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench coding average	73.25	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench mathematics average	88.25	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench data analysis average	70.73	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench language average	73.72	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench if average	80.74	LiveBench	Score for 'Claude 4 Opus Thinking' variant.
livebench global average	71.52	LiveBench	Score for 'Claude 4 Opus' variant.
livebench reasoning average	56.44	LiveBench	Score for 'Claude 4 Opus' variant.
livebench coding average	72.87	LiveBench	Score for 'Claude 4 Opus' variant.
livebench mathematics average	78.79	LiveBench	Score for 'Claude 4 Opus' variant.
livebench data analysis average	66.51	LiveBench	Score for 'Claude 4 Opus' variant.
livebench language average	76.11	LiveBench	Score for 'Claude 4 Opus' variant.
livebench if average	78.38	LiveBench	Score for 'Claude 4 Opus' variant.

Notes

Our most powerful and capable model. Sets new standards in complex reasoning and advanced coding. Excels in top-tier results in reasoning, coding, multilingual tasks, long-context handling, honesty, and image processing. Ideal for applications that require rich, human-like interactions.