‹ PublicAI Index
The LLM benchmark aggregator.
Grok 4.7
SpaceXAI
Strongest in Safety (#4 of 327, on 1 of its 3 boards), weakest in Reasoning (#102 of 195). Above par in 25 of 31 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5.1 and Claude Fable 5 and ahead of GPT-6 Sol and GPT-5.
§ 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.
Safety60−1.4#4/3271/3
Fairness70leads#4/2901/2
Secure code64.9−1.3#13/2681/1
Jailbreak resistance61.8−3.2#25/2661/1
Toxicity avoidance57.1−1.9#45/2661/1
Harm refusal55.5−6#63/2901/2
Agents57.9−8.4#29/2983/6
Web research58.3−5.3#7/361/2
Knowledge work67.3−3.4#9/2082/2
Long-horizon work53.6−20.2#13/321/1
Computer use46.6−21.5#45/691/2
Professional54.7−7.7#34/1942/2
Medical59.3−7.2#14/1491/1
Document design52.7−15.7#15/321/1
Biology research49.5−21.2#17/311/1
Legal59−6.9#19/1731/1
IT operations42−31.8#20/331/1
Cybersecurity54.1−9.8#26/471/1
Finance53−9.4#74/1621/1
Coding55.1−11#34/1794/6
Agentic coding55.3−11.4#37/1734/5
Core abilities55.5−10.1#40/2003/4
Instruction following63−10.8#10/621/1
Data analysis53.9−12.2#16/321/1
Language52.7−13.9#28/621/1
General intelligence53.9−15.1#59/1802/3
Human preference60.7−8.1#79/3541/1
Human preference60.7−8.1#79/3541/1
Reasoning49.7−15.7#102/1954/5
Reasoning49.5−19.5#68/1472/3
Mathematics49.9−13.9#87/1522/2
§ 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 Grok 4.7, left for the other.
§ 3 · Sources
Where the numbers come from
10 publications, 40 figures. Every one links to the page it was read from.
LMArena Text 1442
Artificial Analysis Intelligence Index 35
GDPval-AA 1717
AA-Briefcase 1636
Terminal-Bench 38.8% (Grok Build)
ARC-AGI-2 61.4%
LiveBench 67.9LiveBench · Agentic 45.3LiveBench · Reasoning 82.7LiveBench · Coding 73.1LiveBench · Browser Use 49.3LiveBench · Data Analysis 76.9LiveBench · Research 89.5LiveBench · Design 75
LiveBench 1.0 77.4LiveBench 1.0 · Reasoning 82.7LiveBench 1.0 · Coding 77.2LiveBench 1.0 · Agentic Coding 54LiveBench 1.0 · Mathematics 95.7LiveBench 1.0 · Data Analysis 76.9LiveBench 1.0 · Language 80.1LiveBench 1.0 · Instruction Following 75.3
Vals · Legal Research Bench 47.12%Vals · LegalBench 84.39%Vals · Harvey Legal Agent Benchmark 18.33%Vals · Finance Agent 52.25%Vals · MedCode 49.55%Vals · MedScribe 89.38%Vals · BioMysteryBench 69.26%Vals · CyberBench 69.46%Vals · SRE Bench 3.82%Vals · Vibe Code Bench 86.17% (OpenHands)Vals · Code Migration 44.82%Vals · ProofBench 26%
Enkrypt · Jailbreak risk 2.9%Enkrypt · Harmful content risk 0%Enkrypt · CBRN risk 18.3%Enkrypt · Toxicity risk 1.3%Enkrypt · Bias risk 26.6%Enkrypt · Insecure code risk 2.7%
Badge
[](https://publicai.io/model-index/m/grok-4-7)