evals.report
BenchmarksLabsCompareRun guidesIn the wild
MetaLlama

Llama 4 Maverick

Meta · Llama. Released Apr 5, 2025.

Llama 4 Maverick is a model from Meta in the Llama family, released Apr 5, 2025. evals.report tracks 28 reported Llama 4 Maverick benchmark scores across Berkeley Function Calling Leaderboard, SWE-bench Pro, GPQA Diamond, FrontierMath, AIME (OTIS Mock), LiveCodeBench Pro, WeirdML, Artificial Analysis Intelligence Index, and 20 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

Open28 results

Benchmark results 28

Compare this model
BenchmarkCategoryScoreMetricStatusDate
Berkeley Function Calling LeaderboardTool use37.29%accuracyOfficialApr 5, 2025Details
SWE-bench ProCoding5.24%% resolvedOfficialApr 5, 2025Details
GPQA DiamondReasoning67.0%accuracyOfficialApr 5, 2025Details
FrontierMathReasoning0.69%accuracyOfficialApr 5, 2025Details
AIME (OTIS Mock)Reasoning20.6%accuracyOfficialApr 5, 2025Details
LiveCodeBench ProCoding528Codeforces EloOfficialApr 5, 2025Details
WeirdMLCoding24.5%average accuracyOfficialApr 5, 2025Details
Artificial Analysis Intelligence IndexReasoning18.4IndexUnverifiedApr 5, 2025Details
Epoch Capabilities IndexReasoning133.1IndexOfficialApr 5, 2025Details
Aider PolyglotCoding15.6%% correctOfficialApr 5, 2025Details
MMLU-ProReasoning80.9%accuracyVerifiedApr 5, 2025Details
GAIA: A Benchmark for General AI AssistantsAgents28.6%accuracyUnverifiedApr 5, 2025Details
MathVistaMultimodal73.7%accuracyVerifiedApr 5, 2025Details
GDPvalAgents435EloOfficialApr 5, 2025Details
LiveCodeBenchCoding39.7%Pass@1UnverifiedApr 5, 2025Details
SciCodeCoding33.1%accuracyUnverifiedApr 5, 2025Details
MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark)Multimodal73.4%accuracyUnverifiedApr 5, 2025Details
Global-MMLUReasoning82.5%accuracyUnverifiedApr 5, 2025Details
Arena-Hard-Auto v2.0Chat preference17.2%% win rateOfficialApr 5, 2025Details
EQ-Bench Creative Writing v3Chat preference927EloVerifiedApr 5, 2025Details
Design ArenaChat preference934EloVerifiedApr 5, 2025Details
MASK (Model Alignment between Statements and Knowledge)Other49.73Honesty scoreVerifiedApr 5, 2025Details
BigCodeBenchCoding29.1%calibrated Pass@1VerifiedApr 5, 2025Details
EnigmaEvalReasoning0.58%accuracyVerifiedApr 5, 2025Details
ZeroBenchMultimodal0.0% (pass@1)accuracyVerifiedApr 5, 2025Details
Vectara Hallucination LeaderboardOther8.2%Hallucination RateOfficialApr 5, 2025Details
PolyMath: Evaluating Mathematical Reasoning in Multilingual ContextsReasoning26.1Difficulty-Weighted Accuracy (DW-ACC)VerifiedApr 5, 2025Details
MultiNRCReasoning8.44%accuracyOfficialApr 5, 2025Details