DeepSeek V3.2 (Reasoning)

deepseek-v3-2-reasoning · 4 samples · combined / raw

Scoring

Choose the quality benchmark and final score formula for this model.

Cost adjustment

Tune formulas that account for benchmark cost.

Score over time

44.8 42.8 #1 · 43.8 · 2026-05-11 23:17:33#2 · 43.8 · 2026-05-12 00:00:09#3 · 43.8 · 2026-05-12 01:00:09#4 · 43.8 · 2026-05-12 02:00:11

Cost over time

$76.7 $74.7 #1 · $75.7 · 2026-05-11 23:17:33#2 · $75.7 · 2026-05-12 00:00:09#3 · $75.7 · 2026-05-12 01:00:09#4 · $75.7 · 2026-05-12 02:00:11
Run?Fetched?Score?Quality?Cost?Intel?Code?Agent?MMMU%?
#4 2026-05-12 02:00:11 43.8 43.8 $75.7 41.7 36.7 52.9 -
#3 2026-05-12 01:00:09 43.8 43.8 $75.7 41.7 36.7 52.9 -
#2 2026-05-12 00:00:09 43.8 43.8 $75.7 41.7 36.7 52.9 -
#1 2026-05-11 23:17:33 43.8 43.8 $75.7 41.7 36.7 52.9 -