‹ PublicAI Index
The LLM benchmark aggregator.
Phi 3 Mini 4k Instruct
Microsoft
Strongest in Jailbreak resistance (#58 of 272), weakest in Human preference (#308 of 342). Above par in 5 of 8 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Opus 4.7 and ahead of Command R and Gemini 3.6 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.
Safety53−8.5#105/3371/3
Jailbreak resistance59−5.9#58/2721/1
Secure code57.3−9.2#93/2741/1
Harm refusal53.5−8.1#98/3001/2
Toxicity avoidance53−6#154/2721/1
Fairness44.2−25.8#211/3001/2
Human preference32.3−35.2#308/3421/1
Human preference32.3−35.2#308/3421/1
§ 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 Phi 3 Mini 4k Instruct, left for the other.
§ 3 · Sources
Where the numbers come from
2 publications, 7 figures. Every one links to the page it was read from.
LMArena Text 1143
Enkrypt · Jailbreak risk 5.2%Enkrypt · Harmful content risk 24.4%Enkrypt · CBRN risk 10.5%Enkrypt · Toxicity risk 4.2%Enkrypt · Bias risk 86.8%Enkrypt · Insecure code risk 18.7%
Badge
[](https://publicai.io/model-index/m/phi-3-mini-4k-instruct)