in the loop

DeepSeek V4 Pro 0813
#29 of 274 scored.

DeepSeekAug 13, 2026Open weights

Capabilities Index

Range 153.5–157.4

155.3

Context

Takes text

1.05M

Input

Per 1M tokens

$0.66

Output

Per 1M tokens

$1.98

Benchmarks

How it scores.

  • GPQA Diamond88.9%

    25th of 186

  • FrontierMath64.6%

    31st of 81

  • ARC-AGI-261.3%

    23rd of 79

  • SimpleQA Verified52.9%

    21st of 77

  • Mock AIME98.6%

    19th of 176

Everything else Epoch has for it

  • ARC-AGI90.5%
  • DTBench89.8%
  • WeirdML66.2%
  • LMCA53.5%
  • ProofBench50.0%
  • APEX-Agents47.3%
  • Chess Puzzles44.2%
  • Mystery Game Puzzles37.2%
  • FrontierCode28.5%
  • FrontierMath-Tier-4-v2-Private26.8%

On the index

Around it.

  1. #26Claude Sonnet 5Anthropic · $2.00 in156.2
  2. #27GLM-5.3Z.ai · $0.039 in155.6
  3. #28GPT-5.2 ProOpenAI · $21 in155.4
  4. #29DeepSeek V4 Pro 0813DeepSeek · $0.66 in155.3
  5. #30Claude Opus 4.6Anthropic · $5.00 in155.2
  6. #31Qwen3.8 Max (0902)Alibaba · $2.00 in155.1
  7. #32DeepSeek V4.1 FlashDeepSeek · $0.30 in154.9

Sources