Anthropic · Claude

Claude Opus 4.7

Legacy · Released: April 16, 2026

Context window
1.000.000 tokens
Max output
32.000 tokens
Price input / 1M tokens
$5.00
Price output / 1M tokens
$25.00
Price cache-read / 1M
$0.50

Capabilities

  • Vision
  • Thinking
  • Tool use
  • Computer Use
  • Cache
  • Batch

Benchmark scores

MMMLU
91.5
MMLU-Pro
89.9
HumanEval
96.8
GPQA Diamond
94.2
SWE-bench Pro
64.3
SWE-bench Verified
87.6
Terminal-Bench 2.0
69.4

What do these benchmarks mean? →

Notes

Anthropic flagship model. Sharpest vision (up to 3.75 megapixels) and new xhigh effort setting between high and max. Built-in cybersecurity abuse blocking. New tokenizer can produce up to 35% more tokens per input than 4.6.