‹ PublicAI Index
The LLM benchmark aggregator.
Gemma 4 31B
Google · 31B · open weights
Strongest in Data analysis (#30 of 31), weakest in Agents (#298 of 298). Above par in 11 of 35 scopes. Among the models it meets almost everywhere, it finishes behind Grok 4.6 and DeepSeek V4.1 Flash and ahead of DeepSeek R1.
§ 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.
Knowledge53.5−9#52/1971/3
Document parsing53.7−9.6#39/831/1
Factuality39.4−21.4—/0✱✱
Human preference61.7−7.1#64/3541/1
Human preference61.7−7.1#64/3541/1
Professional43.8−18.6#157/1942/2
Document design26−42.7#30/311/1
Finance50.6−11.8#95/1621/1
Legal43.4−22.5#135/1731/1
Coding39.7−26.5#170/1791/6
Agentic coding38.7−28.2#167/1731/5
Code generation55.9−9.4—/20✱0/1
Reasoning38.7−26.7#176/1951/5
Reasoning31.8−37.2#147/1471/3
Science49.7−13.8—/122✱0/1
Expert reasoning44−22.4—/0✱✱
Safety49.7−11.7#185/3292/3
Factual grounding56.4−14.3#35/1041/1
Secure code60.4−5.8#63/2701/1
Toxicity avoidance52.6−6.4#156/2681/1
Harm refusal48.7−12.8#172/2921/2
Jailbreak resistance42.1−22.9#208/2681/1
Fairness41.2−28.8#261/2921/2
Core abilities37.7−27.6#197/1993/4
Data analysis26−39.2#30/311/1
General intelligence40.2−29.2#149/1783/3
Long context49.8−14.8—/0✱✱
Agents35.4−31#298/2983/6
Long-horizon work34.2−39.8#31/311/1
Web research30−33.9#35/351/2
Computer use30−38.1#68/681/2
Knowledge work41.2−29.5#145/2082/2
Tool use45.8−28.2—/81✱0/1
Decisions60.2−5.7—/234✱0/1
Routing & classification63.7−8—/234✱0/1
§ 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 Gemma 4 31B, left for the other.
§ 3 · Sources
Where the numbers come from
12 publications, 38 figures. Every one links to the page it was read from.
LMArena Text 1452
Artificial Analysis Intelligence Index 4
GDPval-AA 627
AA-Briefcase 369
LiveBench 32.4LiveBench · Agentic 8.5LiveBench · Reasoning 68.3LiveBench · Coding 20.3LiveBench · Browser Use 0LiveBench · Data Analysis 53.3LiveBench · Research 38.1LiveBench · Design 56.5
Kagi LLM Benchmark 63.5%
Vals · CaseLaw 52.63%Vals · MortgageTax 61.37%
ParseBench · Overall 62.4%ParseBench · Tables 80.6%ParseBench · Charts 15%ParseBench · Content faithfulness 89.9%ParseBench · Semantic formatting 69.3%ParseBench · Visual grounding 57.4%
Enkrypt · Jailbreak risk 19.4%Enkrypt · Harmful content risk 6.1%Enkrypt · CBRN risk 33.3%Enkrypt · Toxicity risk 4.4%Enkrypt · Bias risk 91.5%Enkrypt · Insecure code risk 12%
Vectara · Factual consistency 92.6%
tau3-Banking 14.8%Terminal-Bench 2.1 43.4%SciCode 43.4%Humanity's Last Exam (without tools) 23.6%GPQA Diamond 85.7%CritPt 1.4%AA-LCR 68.3%AA-Omniscience Accuracy 20%AA-Omniscience Non-Hallucination 15%
Decision correct 77%
Badge
[](https://publicai.io/model-index/m/gemma-4-31b)