‹ PublicAI Index
The LLM benchmark aggregator.
DeepSeek R1
DeepSeek · 685B · open weights
Strongest in Safe-prompt compliance (#1 of 82), weakest in Agents (#279 of 298). Above par in 14 of 26 scopes. Among the models it meets almost everywhere, it finishes behind GPT-6 Astra and Claude Fable 5 and ahead of Qwen3 235B A22B Instruct and Grok 4.
§ 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.
Coding57.9−8.3#19/1792/6
Code generation60.8−4.5#4/201/1
Agentic coding56.5−10.4#28/1731/5
Core abilities55.7−9.6#37/1991/4
General intelligence58.2−11.2#34/1781/3
Knowledge51−11.5#100/1971/3
Academic knowledge52.4−11.5#70/1231/1
Human preference58.7−10.1#110/3541/1
Human preference58.7−10.1#110/3541/1
Professional48.4−14#119/1941/2
Medical53.9−12.6#56/1491/1
Finance49.1−13.3#102/1621/1
Legal41.7−24.2#146/1731/1
Reasoning46.6−18.8#126/1953/5
Mathematics52−11.8#75/1521/2
Reasoning42−27#110/1472/3
Safety45.2−16.2#278/3293/3
Safe-prompt compliance60.4leads#1/821/1
Fairness55.2−14.8#68/2922/2
Factual grounding46.5−24.2#73/1041/1
Toxicity avoidance55.9−3.1#75/2681/1
Jailbreak resistance32.8−32.2#242/2681/1
Secure code30.1−36.1#247/2701/1
Harm refusal41.4−20.1#276/2922/2
Agents40.7−25.7#279/2982/6
Knowledge work35−35.7#184/2082/2
§ 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 R1, left for the other.
§ 3 · Sources
Where the numbers come from
12 publications, 26 figures. Every one links to the page it was read from.
LMArena Text 1421
GDPval-AA 302
AA-Briefcase 129
ARC-AGI-2 1.1%
LiveCodeBench 73.1%
Aider polyglot 71.4%
Kagi LLM Benchmark 69.4%
SimpleBench 40.8%
Vals · LegalBench 67.32%Vals · CorpFin 54.12%Vals · TaxEval 72.28%Vals · MedQA 90.8%Vals · MMLU Pro 83.18%Vals · AIME 73.96%
HELM Safety · HarmBench 54.6%HELM Safety · SimpleSafetyTests 98.3%HELM Safety · Anthropic Red Team 99.1%HELM Safety · BBQ 96.5%HELM Safety · XSTest 98.8%
Enkrypt · Jailbreak risk 27.2%Enkrypt · Harmful content risk 57.8%Enkrypt · CBRN risk 53.5%Enkrypt · Toxicity risk 2.1%Enkrypt · Bias risk 74.9%Enkrypt · Insecure code risk 73.8%
Vectara · Factual consistency 88.7%
Badge
[](https://publicai.io/model-index/m/deepseek-r1)