evals.report
BenchmarksLabsCompareRun guidesIn the wild
xAIGrok

Grok 4.3

xAI · Grok. Released Apr 17, 2026.

Grok 4.3 is a model from xAI in the Grok family, released Apr 17, 2026. evals.report tracks 12 reported Grok 4.3 benchmark scores across Humanity's Last Exam, LMArena, WeirdML, Artificial Analysis Intelligence Index, τ²-bench (Telecom), GDPval, AA-Omniscience: Knowledge and Hallucination Benchmark, IFBench, and 4 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

12 results

Benchmark results 12

Compare this model
BenchmarkCategoryScoreMetricStatusDate
Humanity's Last ExamReasoning33.12%accuracyOfficialApr 17, 2026Details
LMArenaChat preference1408source-defined ratingOfficialApr 17, 2026Details
WeirdMLCoding49.9%average accuracyOfficialApr 17, 2026Details
Artificial Analysis Intelligence IndexReasoning53.2IndexUnverifiedApr 17, 2026Details
τ²-bench (Telecom)Tool use97.7%pass^1OfficialApr 17, 2026Details
GDPvalAgents1495EloOfficialApr 17, 2026Details
AA-Omniscience: Knowledge and Hallucination BenchmarkReasoning18AA-Omniscience IndexOfficialApr 17, 2026Details
IFBenchReasoning83.3%accuracyOfficialApr 17, 2026Details
WebDev ArenaChat preference1377EloVerifiedApr 17, 2026Details
Search ArenaChat preference1189EloVerifiedApr 17, 2026Details
Design ArenaChat preference1260EloVerifiedApr 17, 2026Details
Vibe Code BenchCoding19.40%Overall accuracyVerifiedApr 17, 2026Details