DeepSeek V3.2 Speciale: benchmark results

DeepSeek V3.2 Speciale variant row. Provider: DeepSeek. Released 2025-12-01. Access: Open.

Unified ELO 1656 ± 1, rank #119 of 1392 rated models, from 113 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
MathArena - CMIMC 202594.38Accuracy (%)100
OpenCompass Research - AIME 202596Score (%)100
UGI Leaderboard65.37UGI Score99.5
JudgeBench Reasoning96.94Accuracy (%)98
OpenCompass Research - HLE28.6Score (%)96.8
BenchTable76.9Total Score (%)94.5
LLM Stats (HMMT 2025)99.2Score (%)93.8
MathArena - HMMT Feb 202597.5Accuracy (%)92.6
MathArena - BRUMO 202599.17Accuracy (%)92.2
OpenCompass Research - GPQA Diamond86.7Score (%)91.5
UGI - Natural Intelligence53.02NatInt Score91
OpenCompass78.2Average Score90.4

Interactive version: theaggregate.ai/model?slug=deepseek-v3-2-speciale · How It Works · Data refreshed daily, snapshot 2026-09-05.