Models · Claude Opus 4
MODEL REGISTRY

Claude Opus 4

Anthropic's most capable model. State-of-the-art on coding benchmarks, extended thinking, and agentic tasks. The first model to sustain peak performance on multi-hour agent workflows.

Anthropic CLOSED RELEASED May 22, 2025 ACTIVE
CONTEXT
200K
tokens
PARAMETERS
---
total
BENCHMARKS
3
tracked
INPUT
$15
per 1M tokens
OUTPUT
$75
per 1M tokens
Specifications
Lab Anthropic
Type CLOSED
Released May 22, 2025
Context Window 200K tokens
Training Cutoff 2025-03
Status ACTIVE
Pricing
INPUT
$15.00
per 1M tokens
OUTPUT
$75.00
per 1M tokens
Benchmarks 3 tracked
Benchmark Score Date
SWE-bench Verified
72.5 2025-05
Terminal-bench
43.2 2025-05
GPQA Diamond
74.9 2025-05
Capabilities 5
textvisionextended-thinkingtool-useagentic
About

Claude Opus 4 is Anthropic’s most intelligent model, built for sustained autonomous work. It set new records on SWE-bench Verified and introduced parallel tool use and extended thinking for complex reasoning chains.