‹ PublicAI IndexGoogle
The LLM benchmark aggregator.
Gemini 2.5 Flash-Lite
Strongest in Factual grounding (#3 of 104), weakest in Human preference (#156 of 354). Above par in 7 of 20 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 4.8 and GPT-5.6 Terra and ahead of O3 Mini and DeepSeek V3.1.
§ 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.
Safety54.4−7#64/3292/3
Factual grounding66.9−3.8#3/1041/1
Safe-prompt compliance58.2−2.2#14/821/1
Fairness55.3−14.7#66/2921/2
Harm refusal50−11.5#153/2921/2
Knowledge48.9−13.6#130/1971/3
Academic knowledge47.4−16.5#92/1231/1
Agents49.7−16.7#133/2981/6
Tool use49.3−24.7#38/811/1
Professional47.2−15.2#135/1941/2
Legal52.9−13#70/1731/1
Finance46.9−15.5#113/1621/1
Medical43.7−22.8#124/1491/1
Reasoning45.5−19.9#135/1951/5
Science45.9−17.6#89/1221/1
Mathematics43.6−20.2#114/1521/2
Core abilities45.8−19.5#135/1991/4
General intelligence44−25.4#126/1781/3
Human preference54.3−14.5#156/3541/1
Human preference54.3−14.5#156/3541/1
§ 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 Gemini 2.5 Flash-Lite, left for the other.
§ 3 · Sources
Where the numbers come from
6 publications, 19 figures. Every one links to the page it was read from.
LMArena Text 1375
BFCL v4 36.87%
Kagi LLM Benchmark 40.5%
Vals · LegalBench 82.04%Vals · CorpFin 57.58%Vals · TaxEval 64.72%Vals · MortgageTax 57.55%Vals · MedQA 88.87%Vals · MedCode 34.19%Vals · MedScribe 66.88%Vals · GPQA Diamond 70.2%Vals · MMLU Pro 79.12%Vals · AIME 42.08%
HELM Safety · HarmBench 67%HELM Safety · SimpleSafetyTests 96.5%HELM Safety · Anthropic Red Team 98.7%HELM Safety · BBQ 94.9%HELM Safety · XSTest 97.8%
Vectara · Factual consistency 96.7%
Badge
[](https://publicai.io/model-index/m/gemini-2-5-flash-lite)