Skip to content
SOLVENCY

Head to head

Claude Opus 5 vs DeepSeek V4 Pro

Claude Opus 5: currentDeepSeek V4 Pro: currentheavy tier

No verdict

Cost per solved task needs a published pass rate for both models. At least one has none, so it is reported as missing rather than estimated.

Source: Artificial Analysis (artificialanalysis.ai) · verified 2026-08-21 · prices from each provider's pricing page

Input $/M tokens

lower is better

  • Claude Opus 5$5
  • DeepSeek V4 Pro$1.32

Output $/M tokens

lower is better

  • Claude Opus 5$25
  • DeepSeek V4 Pro$3.96

Cost per solved task

needs a pass rate for both

  • Claude Opus 5$12.01
  • DeepSeek V4 Promissing
Claude Opus 5DeepSeek V4 Pro
Input $/Mtok$5$1.32
Output $/Mtok$25$3.96
Cached input $/Mtok$0.5$0.044
Pass rate68%missing
Pass rate sourceAAmissing
Cost basismeasuredmissing
Cost per solved task$12.01missing

verified 2026-08-21 · heavy task tier; measured rows ignore the tier· Methodology· Claude Opus 5· DeepSeek V4 Pro