Claude Opus 4.5

Anthropic🇺🇸 United States
active

Anthropic's most capable Claude 4.5 model.

Context window200K tokens
Input / 1M tokens$5
Output / 1M tokens$25

Version History

claude-opus-4-5

Claude Opus 4.5 — improved agentic coding and reasoning.

Benchmark Scores

Full leaderboard →
87.5%
AIME 2024
89.0%
AIME 2025
94.8%
DocVQA
87.0%
GPQA
10.9%
Hallucination (Summarization)
75.4%
Hallucination Rate
73.8%
LiveCodeBench
93.8%
MATH
89.8%
MMLU
83.2%
MMLU-Pro
80.7%
MMMU
63.0 tokens_per_sec
Speed (tok/s)
80.9%
SWE-bench Verified