supermarket

Language

DeepSeek-V3.2

By DeepSeek, in the DeepSeek-V family.

Released
1 Dec 2025
Status
Legacy
Context
131.1K
Max output
—
Input per 1M tokens
—
Output per 1M tokens
—
Open weights
Yes
License
mit
Parameters
671B, 37B active
Knowledge cutoff
—
Reasoning
Yes
Modalities
text → text

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Category scores
CategoryScoreRankCoverage
Knowledge & reasoning21.7752 of 4
Coding & software engineering74.2Not ranked1 of 5
Human preference49.0Not ranked2 of 5

Benchmark results

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning

Knowledge & reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Chess PuzzlesPuzzles solved14.0%Independent—16 Dec 2025Epoch AI
Chess PuzzlesPuzzles solved1.0%Independent—16 Jul 2026Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)85.6%Third party——https://conceptualreasoning.ai/dtbench via Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)62.7%Third party——https://conceptualreasoning.ai/dtbench via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.2882Third party——https://conceptualreasoning.ai/lmca via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.1524Third party——https://conceptualreasoning.ai/lmca via Epoch AI

Scientific reasoning

Scientific reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
GPQA DiamondAccuracy83.4%Independent—16 Dec 2025Epoch AI
GPQA DiamondAccuracy71.2%Independent—16 Jul 2026Epoch AI

Mathematics

Mathematics results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
OTIS Mock AIME 2024-2025Accuracy87.8%Independent—16 Dec 2025Epoch AI
OTIS Mock AIME 2024-2025Accuracy48.9%Independent—16 Jul 2026Epoch AI

Coding & software engineering

Coding & software engineering results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Aider PolyglotPercent correct74.2%Third party——https://aider.chat/docs/leaderboards/ via Epoch AI
Aider PolyglotPercent correct70.2%Third party——https://aider.chat/docs/leaderboards/ via Epoch AI
Terminal-Bench 2.0Accuracy (mean over trials)39.6%Third party——https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI

Long context

Long context results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
CL-benchTasks solved12.4%Source not stated——Epoch AI via Epoch AI
CL-bench LifeTasks solved7.4%Source not stated——Epoch AI via Epoch AI

Human preference

Human preference results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1362Independentthinking25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1325Independent—25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes9388Independent—25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes6922Independentthinking25 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1425Independent—25 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1423Independentthinking25 Sep 2026Arena
Text Arena (arena.ai)Votes48010Independent—25 Sep 2026Arena
Text Arena (arena.ai)Votes41980Independentthinking25 Sep 2026Arena

Other DeepSeek-V versions

Sources