‹ PublicAI Index
The LLM benchmark aggregator.
C4ai Aya Expanse 8B
Cohere · 8B
Strongest in Jailbreak resistance (#33 of 272), weakest in Human preference (#268 of 342). Above par in 6 of 9 scopes. Among the models it meets almost everywhere, it finishes behind Phi 4 and Claude Sonnet 5 and ahead of Kimi K2.6 and Gemma 4 31B.
§ 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.
Safety53.9−7.6#73/3372/3
Jailbreak resistance60.6−4.3#33/2721/1
Toxicity avoidance57.4−1.6#43/2721/1
Factual grounding51.2−19.4#52/1011/1
Harm refusal53.9−7.7#89/3001/2
Secure code57.1−9.4#96/2741/1
Fairness45.4−24.6#187/3001/2
Human preference40−27.5#268/3421/1
Human preference40−27.5#268/3421/1
§ 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 C4ai Aya Expanse 8B, left for the other.
§ 3 · Sources
Where the numbers come from
3 publications, 8 figures. Every one links to the page it was read from.
LMArena Text 1223
Enkrypt · Jailbreak risk 3.8%Enkrypt · Harmful content risk 32.8%Enkrypt · CBRN risk 4.8%Enkrypt · Toxicity risk 1.1%Enkrypt · Bias risk 85%Enkrypt · Insecure code risk 19.1%
Vectara · Factual consistency 90.5%
Badge
[](https://publicai.io/model-index/m/c4ai-aya-expanse-8b)