DeepSeek V3.2

DeepSeek V3.2 is a non-reasoning model from DeepSeek. 26 benchmarks count toward its score, in 6 categories.

availableShows if the model has enough results for an index.
IndexOverall score. 50 is the middle.44.4 ±4.7
CoverageShare of the index weight with results.85%
SpeedOutput tokens per second.33/s
Input / 1MUS dollars per 1M input tokens.$0.269
Output / 1MUS dollars per 1M output tokens.$0.4
ContextMaximum tokens in one request.164K
EloLMArena rating and rank.1425 (#96)

50 is the middle of the board. The range shows the doubt in the index.

46,458 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category. 50 is the middle.

50 is the middle
AgenticMulti-step tasks with tools.
42.7
CodingCode writing and repair.
49.4
ReasoningLogic problems and puzzles.
41.8
MultimodalTasks with images and text.
N/A
KnowledgeFacts and expert knowledge.
44.8
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
38.8
MathMath problems.
47.7

Results

26 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result on the index scale.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
τ²-bench TelecomAgentic96.2%67.8DeepSeek2 Mar 2026Sierra Research
MGSMMultilingual90.9%9 Jan 2026Vals AI
MMLU ProKnowledge83.1%51.41 Sept 2026Vals AI
τ²-bench RetailAgentic81.1%57.0DeepSeek30 Apr 2026Sierra Research
τ²-Bench Tool-Agent-User EvaluationAgentic78.9%55.4Victor Barres et al.
GPQA diamondKnowledge77.3%49.6Epoch AI
GPQA DiamondKnowledge76.3%48.71 Sept 2026Vals AI
Artificial Analysis GPQA DiamondKnowledge75.1%45.5Artificial Analysis
React Native EvalsCoding71.5%47.8Callstack
SWE-bench VerifiedCoding70.0%53.7high effort · mini-SWE-agent1 Sept 2026SWE-bench team
LiveCodeBenchCoding69.9%47.81 Sept 2026Vals AI
OTIS Mock AIME 2024-2025Math68.4%49.8Epoch AI
AIMEMath64.8%43.916 Apr 2026Vals AI
τ²-bench AirlineAgentic63.8%44.6DeepSeek2 Mar 2026Sierra Research
SWE-RebenchCoding60.9%Nebius
SWE-bench MultilingualMultilingual59.0%mini-SWE-agent20 Feb 2026SWE-bench team
SWE-bench MultilingualCoding59.0%mini-SWE-agent2 Sept 2026SWE-bench team
ARC-AGI-1 (semi-private)Reasoning57.0%54.1ARC Prize Foundation
Terminal-Bench 1.0Agentic50.0%53.912 Jan 2026Vals AI
Artificial Analysis IFBenchInstruction49.0%38.8Artificial Analysis
Artificial Analysis Long Context ReasoningReasoning45.7%39.9Artificial Analysis
Claw-EvalAgentic40.2%21.7Bowen Ye et al.
Terminal-Bench 2.0Agentic34.8%47.54 Jun 2026Vals AI
Gert Labs Composite Game BenchmarkAgentic29.6%38.9Gert Labs
Artificial Analysis Omniscience AccuracyKnowledge24.0%43.6Artificial Analysis
FrontierMath-2025-02-28-PrivateMath22.1%51.2Epoch AI
VITA-BenchAgentic18.5%38.3Meituan LongCat Team
Artificial Analysis Intelligence IndexKnowledge16.0%42.5Artificial Analysis
IOI v1Coding14.4%48.19 Aug 2026Vals AI
Artificial Analysis Humanity's Last ExamKnowledge11.2%36.7Artificial Analysis
Chess PuzzlesReasoning7.5%33.2Epoch AI
ARC-AGI-2 (semi-private)Reasoning4.0%41.7ARC Prize Foundation
FrontierMath-Tier-4-2025-07-01-PrivateMath2.1%45.7Epoch AI
Critical Physics TasksReasoning0.9%40.3Artificial Analysis

26 benchmarks count, from 30 of 34 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence statedSierra Research, collected directlyMIT — results are in the licensed repositoryVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AIEpoch AI, collected directlyCC BY — free to use and redistribute with attributionSWE-bench team, collected directlyNo licence stated. The repository publishes submission records for reproducibility and transparency and asks that SWE-bench be citedARC Prize Foundation, collected directlyNo licence stated. Their terms ask for written permission before commercial use

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from DeepSeek

DeepSeek V4.1 Flash70.6DeepSeek V4 Pro 081368.8DeepSeek V4 Flash 073164.6DeepSeek V4 Pro 042362.4DeepSeek V4 Flash 042363.0