DeepSeek V3.2 Exp — Benchmarks

Benchmark scores for DeepSeek V3.2 Exp aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

No overall rank for DeepSeek V3.2 Exp: an overall score is only published when a model has been measured on enough of the benchmarks current models are still submitted to. Its individual scores below stand on their own — see the methodology for how the overall score is built.

Coding

BenchmarkScoreOpen rankAll models
SciCode38.9%#29 / 62#121 / 161
Aider Polyglot74.2%#1 / 18#14 / 69
WeirdML39.5%#24 / 46#109 / 161
LMArena WebDev1271.8#42 / 48#106 / 123

Instruction Following

BenchmarkScoreOpen rankAll models
LMArena Text1425.3#26 / 182#107 / 391

Reasoning

BenchmarkScoreOpen rankAll models
DTBench87.7%#10 / 67#77 / 210
LMCA29.1%#24 / 53#118 / 172
CritPt2.9%#23 / 62#98 / 172
Fiction.LiveBench83.3%#2 / 24#9 / 58
CL-bench13.2%#5 / 7#20 / 22
CL-bench Life9.5%#4 / 6#14 / 17

Scores aggregated from public benchmark sources (each linked from the benchmark pages). llmrun does not run these benchmarks.