latentbrief

Anthropic · Claude 4

Claude Opus 4.8

Anthropic's heavyweight for hard reasoning and agentic work.

Opus is the Claude you reach for when output quality buys back its premium - long agent runs, hard reasoning, work where a single dropped step costs more than the token bill. It pairs Anthropic's discipline on tool use with extended thinking that genuinely helps on the hardest problems. Sonnet is the right default; Opus is the right answer when Sonnet starts to slip.

Specs

Context window
1M tokens
Input ($/1M tokens)
$5.00
Output ($/1M tokens)
$25.00
Modalities
Text · Image
Released
May 28, 2026
Weights
Closed

Capabilities

  • Tool use
  • Vision
  • Extended thinking
  • Prompt caching
  • Open weights

API model id

claude-opus-4-8

Vendor docs: docs.anthropic.com

Compare Claude Opus 4.8 with