‹ PublicAI Index
The LLM benchmark aggregator.
O3 Mini
OpenAI
Strongest in Code generation (#23 of 77, on 1 of its 2 boards), weakest in Agents (#243 of 268). Above par in 13 of 24 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of GPT-5.4 Mini and O1 Mini.
§ 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.
Coding54.8−14.3#32/1652/5
Code generation56−14#23/771/2
Agentic coding54.1−14.6#46/1571/4
Safety55.4−6.1#45/3372/3
Harm refusal57.3−4.3#29/3002/2
Jailbreak resistance60.3−4.6#40/2721/1
Safe-prompt compliance49.9−10.5#56/821/1
Secure code58.2−8.3#81/2741/1
Fairness51.2−18.8#99/3002/2
Toxicity avoidance52.4−6.6#169/2721/1
Knowledge47.4−16.1#102/1381/2
Academic knowledge46.8−17.1#94/1231/1
Professional47.2−14.4#112/1681/1
Medical56−10.8#29/1401/1
Legal44.9−22.2#104/1511/1
Finance42.8−20.1#122/1511/1
Reasoning46.2−20.7#114/1783/4
Mathematics55.4−10.3#43/1401/2
Science49.6−13.9#77/1221/1
Reasoning37.3−31.8#128/1402/3
Human preference53.5−14#155/3421/1
Human preference53.5−14#155/3421/1
Agents39.4−28.7#243/2682/5
Knowledge work36.3−37.3#155/1782/2
§ 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 O3 Mini, left for the other.
§ 3 · Sources
Where the numbers come from
10 publications, 25 figures. Every one links to the page it was read from.
LMArena Text 1364
GDPval-AA 218
AA-Briefcase 143
ARC-AGI-2 3%
LiveCodeBench 67.4%
Aider polyglot 60.4%
SimpleBench 22.8%
Vals · LegalBench 71.54%Vals · CorpFin 45.3%Vals · TaxEval 69.42%Vals · MedQA 94.83%Vals · GPQA Diamond 75.5%Vals · MMLU Pro 78.69%Vals · AIME 86.46%
HELM Safety · HarmBench 95.2%HELM Safety · SimpleSafetyTests 99%HELM Safety · Anthropic Red Team 98.6%HELM Safety · BBQ 94.1%HELM Safety · XSTest 94.1%
Enkrypt · Jailbreak risk 4.1%Enkrypt · Harmful content risk 15.6%Enkrypt · CBRN risk 8.3%Enkrypt · Toxicity risk 4.6%Enkrypt · Bias risk 81.9%Enkrypt · Insecure code risk 16.9%
Badge
[](https://publicai.io/model-index/m/o3-mini)