Benchfolio

Leaderboard / Model

DeepSeek V4 Pro 0813

DeepSeek · Open weights · Released Aug 13, 2026 · Price not listed

45Overall · rank 22

Score by job

Coding

Not enough data · 2 of 5 benchmarks

BenchmarkResultVariantz
FrontierCode28.5%deepseek-v4-pro-0813_high-1.14
LMArena Coding1467 ratingdeepseek-v4-pro-high-20260813-1.13

Agents

Not enough data · 1 of 4 benchmarks

BenchmarkResultVariantz
APEX-Agents47.3%deepseek-v4-pro-0813_unknown-0.48

Reasoning

48 · rank 22 · 2 of 4 benchmarks

BenchmarkResultVariantz
GPQA Diamond91.7%deepseek-v4-pro-0813_max0.04
ARC-AGI-261.3%deepseek-v4-pro-0813_max-0.13

Math

37 · rank 23 · 5 of 5 benchmarks

BenchmarkResultVariantz
FrontierMath Tiers 1–364.6%deepseek-v4-pro-0813_max-0.24
FrontierMath Tier 426.8%deepseek-v4-pro-0813_max-0.59
OTIS Mock AIME98.6%deepseek-v4-pro-0813_max0.75
ProofBench50.0%deepseek-v4-pro-0813_unknown0.09
LMArena Math1446 ratingdeepseek-v4-pro-high-20260813-1.28

Writing

49 · rank 32 · 2 of 2 benchmarks

BenchmarkResultVariantz
LMArena Creative Writing1435 ratingdeepseek-v4-pro-high-20260813-0.49
LMArena Instruction Following1443 ratingdeepseek-v4-pro-high-20260813-0.82

Design

47 · rank 18 · 1 of 1 benchmarks

BenchmarkResultVariantz
LMArena WebDev1581 ratingdeepseek-v4-pro-high-202608130.22