‹ PublicAI Index
The LLM benchmark aggregator.
Grok 4.3
SpaceXAI
Strongest in Science (#22 of 122), weakest in Safety (#247 of 329). Above par in 8 of 24 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5 and Qwen3.8 Max and ahead of GLM-4.6 and Kimi K2.
§ 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.
Human preference60.7−8.1#78/3541/1
Human preference60.7−8.1#78/3541/1
Knowledge52.4−10.1#80/1971/3
Academic knowledge55.8−8.1#52/1231/1
Reasoning50.5−14.9#94/1952/5
Science60.7−2.8#22/1221/1
Mathematics43.5−20.3#115/1521/2
Professional50−12.4#97/1941/2
Legal55.7−10.2#41/1731/1
Finance48−14.4#110/1621/1
Medical45.4−21.1#117/1491/1
Agents49.4−17#136/2982/6
Knowledge work49.1−21.6#102/2082/2
Core abilities44.5−20.8#151/1991/4
Instruction following40.8−33#49/621/1
Language44.1−22.5#50/621/1
Coding38.7−27.5#175/1792/6
Agentic coding38.2−28.7#169/1732/5
Safety47.1−14.3#247/3291/3
Secure code55.8−10.4#110/2701/1
Harm refusal48.4−13.1#179/2921/2
Fairness44−26#212/2921/2
Toxicity avoidance46.4−12.6#220/2681/1
Jailbreak resistance38.6−26.4#221/2681/1
§ 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 Grok 4.3, left for the other.
§ 3 · Sources
Where the numbers come from
6 publications, 31 figures. Every one links to the page it was read from.
LMArena Text 1442
GDPval-AA 946
AA-Briefcase 763
LiveBench 1.0 62.3LiveBench 1.0 · Reasoning 70.8LiveBench 1.0 · Coding 69.9LiveBench 1.0 · Agentic Coding 18.5LiveBench 1.0 · Mathematics 84.3LiveBench 1.0 · Data Analysis 55.8LiveBench 1.0 · Language 73.6LiveBench 1.0 · Instruction Following 62.8
Vals · Legal Research Bench 15.38%Vals · CaseLaw 79.31%Vals · LegalBench 84.46%Vals · Finance Agent 37.73%Vals · CorpFin 68.53%Vals · TaxEval 70.81%Vals · MortgageTax 48.25%Vals · MedCode 38.07%Vals · MedScribe 74.4%Vals · SWE-bench Verified 71.4% (Mini-SWE-agent)Vals · Vibe Code Bench 19.4% (OpenHands)Vals · Code Migration 6.79%Vals · GPQA Diamond 91.41%Vals · MMLU Pro 85.84%
Enkrypt · Jailbreak risk 22.4%Enkrypt · Harmful content risk 5%Enkrypt · CBRN risk 34.8%Enkrypt · Toxicity risk 8.6%Enkrypt · Bias risk 87.1%Enkrypt · Insecure code risk 21.3%
Badge
[](https://publicai.io/model-index/m/grok-4-3)