Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

DeepSeek V3.2

DeepSeek · 685B · open weights

Strongest in Tool use (#11 of 81), weakest in Harm refusal (#254 of 300). Above par in 14 of 26 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5 and Claude Opus 5.5 and ahead of Kimi K2 and Gemini 2.5 Flash.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Agents55#64/268
Tool use63.7#11/81
Knowledge work49.4#86/178
Knowledge53.8#64/138
Academic knowledge54.6#59/123
Reasoning51#82/178
Mathematics54.9#53/140
Science52.9#68/122
Reasoning43.6#90/140
Coding48.8#83/165
Agentic coding48.6#79/157
Human preference59.4#97/342
Human preference59.4#97/342
Core abilities49.8#99/204
General intelligence49.7#100/204
Professional48.2#102/168
Medical55.6#34/140
Legal46.4#97/151
Finance45.3#113/151
Safety48.9#213/337
Factual grounding59.3#26/101
Secure code57.6#91/274
Fairness50.5#105/300
Toxicity avoidance51.1#187/272
Jailbreak resistance36.8#230/272
Harm refusal44.6#254/300

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for DeepSeek V3.2, left for the other.

Command R Pluswon 15 · lost 2
Mistral Medium 3.5won 15 · lost 2
Command Awon 21 · lost 3
Mistral Smallwon 18 · lost 3
Mistral Large 3won 17 · lost 3
Llama 4 Scout Instructwon 20 · lost 4
GPT-4o Miniwon 19 · lost 4
Gemini 2.0 Flashwon 18 · lost 4
GPT-4.1 Nanowon 20 · lost 5
GPT OSS 20Bwon 16 · lost 5
Gemini 1.5 Flashwon 12 · lost 4
Gemma 3 27B Itwon 12 · lost 4
Claude 3.5 Haikuwon 15 · lost 5
Ling 3.0 Flashwon 13 · lost 5
Gemini 1.5 Prowon 13 · lost 5
GPT-5.4 Nanowon 14 · lost 6
Claude Haiku 4.5won 18 · lost 8
Nemotron 3 Ultrawon 11 · lost 5
Mistral Largewon 13 · lost 6
Gemini 2.5 Flash Litewon 13 · lost 6
GPT OSS 120Bwon 17 · lost 8
GPT-4.1 Miniwon 17 · lost 8
Grok 3won 14 · lost 7
DeepSeek R1won 16 · lost 8
DeepSeek V3won 16 · lost 9
GPT-5.4 Miniwon 12 · lost 7
GLM-4.6won 15 · lost 10
GPT-5 Nanowon 13 · lost 10
GLM-4.7won 9 · lost 7
MiniMax M3won 14 · lost 11
Qwen3.6 Pluswon 10 · lost 8
Gemini 3.5 Flash Litewon 10 · lost 8
O3 Miniwon 12 · lost 10
Grok 4.3won 13 · lost 11
GPT-4owon 11 · lost 10
Claude 3.7 Sonnetwon 11 · lost 10
Grok 4.1 Fastwon 12 · lost 11
Gemini 2.5 Flashwon 13 · lost 12
Kimi K2won 13 · lost 12
Inkling Smallwon 8 · lost 8
Grok 4 Fastwon 9 · lost 9
O4 Miniwon 12 · lost 13
Mimo V2.5 Prowon 11 · lost 12
MiniMax M2.7won 10 · lost 11
Claude 3.5 Sonnetwon 9 · lost 10
GLM-4.5won 9 · lost 10
Gemini 3.1 Flash Litewon 9 · lost 10
Qwen3 235B A22B Instructwon 11 · lost 13
Gemma 4 31Bwon 10 · lost 12
Claude Sonnet 4won 11 · lost 14
GPT-4.1won 11 · lost 14
MiniMax M2.5won 7 · lost 9
GPT-5 Miniwon 11 · lost 15
Kimi K2 Thinkingwon 8 · lost 12
Inklingwon 9 · lost 16
Gemini 2.5 Prowon 9 · lost 16
Kimi K2.6won 9 · lost 16
O1won 7 · lost 13
Grok 4won 8 · lost 16
Qwen3.7 Maxwon 8 · lost 16
Kimi K2.5won 8 · lost 17
GLM-5.2won 8 · lost 17
Claude Sonnet 4.6won 7 · lost 16
Gemini 3.1 Prowon 6 · lost 14
DeepSeek V4 Flashwon 7 · lost 17
Gemini 3.5 Flashwon 7 · lost 17
GPT-5.6 Lunawon 7 · lost 18
Qwen3.8 27Bwon 5 · lost 14
Claude Opus 4won 6 · lost 17
Gemini 3 Flashwon 4 · lost 12
GLM-5.3 Flashwon 6 · lost 18
Gemini 3.6 Flashwon 6 · lost 18
DeepSeek V4.1 Flashwon 4 · lost 13
DeepSeek V4 Prowon 5 · lost 20
GPT-6 Lunawon 4 · lost 17
GPT-5.1won 4 · lost 19
GLM-5won 3 · lost 15
Grok 4.5won 4 · lost 20
Mimo V2.6 Prowon 3 · lost 16
Claude Sonnet 4.5won 4 · lost 22
Claude Opus 4.5won 3 · lost 17
Claude Opus 4.6won 3 · lost 18
GPT-6 Solwon 3 · lost 19
GLM-5.1won 3 · lost 19
Gemini 3 Prowon 3 · lost 19
Qwen3.8 Maxwon 3 · lost 21
Muse Spark 1.1won 3 · lost 21
Claude Sonnet 5won 3 · lost 22
GPT-5.6 Terrawon 3 · lost 22
GPT-5.4won 3 · lost 22
GPT-5.6 Solwon 3 · lost 22
GLM-5.3won 2 · lost 16
Grok 4.7won 2 · lost 19
O3won 2 · lost 22
GPT-5won 2 · lost 23
GPT-5.2won 2 · lost 23
GPT-5.5won 2 · lost 23
GPT-6 Astrawon 1 · lost 16
Muse Spark 1.2won 1 · lost 16
Kimi K3won 1 · lost 17
Grok 4.6won 1 · lost 23
Claude Opus 4.8won 1 · lost 23
Gemini 3.8 Flashwon 1 · lost 23
Claude Opus 4.7won 1 · lost 24
Gemini 3.7 Flashwon 0 · lost 18
Claude Fable 5.1won 0 · lost 18
Claude Fable 5won 0 · lost 18
Claude Opus 5.5won 0 · lost 21
Claude Opus 5won 0 · lost 24

§ 3 · Sources

Where the numbers come from

9 publications, 23 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1425
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 681
ARC-AGI-2 ↗1 measureread 2026-09-26
ARC-AGI-2 4%
Aider polyglot ↗1 measureread 2026-09-26
Aider polyglot 74.2%
BFCL ↗1 measureread 2026-09-26
BFCL v4 56.73%
Kagi LLM Benchmark ↗1 measureread 2026-09-26
Kagi LLM Benchmark 52.2%
Vals.ai ↗10 measuresread 2026-09-26
Vals · CaseLaw 55.41%Vals · LegalBench 76.08%Vals · CorpFin 50.97%Vals · TaxEval 68.15%Vals · MedQA 93.92%Vals · SWE-bench Verified 67.6% (Mini-SWE-agent)Vals · Vibe Code Bench 5.11% (OpenHands)Vals · GPQA Diamond 80.3%Vals · MMLU Pro 84.92%Vals · AIME 84.58%
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
Enkrypt · Jailbreak risk 23.9%Enkrypt · Harmful content risk 53.3%Enkrypt · CBRN risk 18.5%Enkrypt · Toxicity risk 5.5%Enkrypt · Bias risk 77%Enkrypt · Insecure code risk 18.2%
Vectara · Factual consistency 93.7%

Badge

PublicAI Index badge for DeepSeek V3.2[![PublicAI Index](https://publicai.io/model-index/badge?model=deepseek-v3-2)](https://publicai.io/model-index/m/deepseek-v3-2)