AI War Tracker
← All models

Claude Opus 4.8 vs DeepSeek-V3.2

Anthropic vs DeepSeek — benchmarks, pricing, and capabilities side by side.

  • Claude Opus 4.8 has the higher intelligence index (73.1 vs 71.7)
  • DeepSeek-V3.2 is cheaper ($0.25 vs $6.30 per 1M input)
Claude Opus 4.8DeepSeek-V3.2
Intelligence index73.171.7
DeveloperAnthropicDeepSeek
TypeLLMLLM
AccessOpen weights
Context window131,072 tokens
Input price$6.30 / 1M$0.25 / 1M
Output price$25.00 / 1M$0.38 / 1M
Speed66 tok/s
ReleasedMay 28, 2026December 1, 2025
Parameters671B (37B active)
Input modalitiesText
Output modalitiesText

Shared benchmarks

Claude Opus 4.8
DeepSeek-V3.2
AA-LCR
67.7
65
GPQA Diamond
92
84
Humanity’s Last Exam
45.7
22.2
SciCode
53.5
38.9
Terminal-Bench
58.3
35.6
τ²-bench
94.4
90.6