Grok 3

Grok 3 places in the top three on 5 benchmarks.

xAI Released April 9th, 2025 Family Grok 3 The lab's model card Compare with others

Benchmarks with a score

Ranked first, out of five or more models

In the top three

xAI Apr 2025

Grok 3

Closed weights Flagship

137.5 #133 of 256

Epoch Capabilities Index

Self-Reported Index 142.0 #164 of 404

Agents' Last Exam 4.61% #31 /31
GPQA Diamond 75.8% #109 /229
SimpleQA Verified 39.3% #23 /56

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
Vectara Hallucination Leaderboard (HHEM)	2.10%	#26 of 148	Official leaderboard
IFEval	88.4%	#12 of 60	Official leaderboard
Vectara Hallucination Leaderboard (2025 dataset)	5.80%	#23 of 104	Official leaderboard
ECLeKTic	29%	#8 of 35	Official leaderboard
LOFT	83.3%	#2 of 8	Lab model card
LLM Confabulation Leaderboard	14.2%	#17 of 58	Official leaderboard
FaithJudge	15.3%	#11 of 37	Official leaderboard
SimpleQA	43.6%	#18 of 60	Lab model card
MARS-Bench	51.2%	#5 of 15	Official leaderboard
LMArena (Elo)	1,411	#118 of 335	Official leaderboard
SimpleQA Verified	39.3%	#23 of 56	Official leaderboard
HUMAINE	29	#21 of 50	Official leaderboard
MMLU-Pro	79.9%	#81 of 167	Lab model card
WritingBench	71.7	#24 of 45	Official leaderboard
EQ-Bench 3	1,165	#46 of 78	Official leaderboard
Fiction.LiveBench	58.3%	#35 of 57	Fiction.live leaderboard, via Epoch AI
FACTS Grounding	75.7%	#21 of 34	Official leaderboard
Creative Writing v3	62.2	#78 of 120	Official leaderboard
SnorkelSpatial	12.7%	#24 of 27	Official leaderboard
Phare	53.5%	#62 of 62	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
AIME 2024	93.3%	#5 of 88	Lab model card
MATH Level 5	88.8%	#19 of 94	Epoch AI run
AIME 2025	93.3%	#31 of 143	Lab model card
Misguided Attention	55.1%	#14 of 45	Official leaderboard
GPQA Main	65%	#20 of 60	Official leaderboard
SnorkelGraph	64%	#9 of 24	Official leaderboard
PolyMath	35.2%	#16 of 38	Official leaderboard
GPQA Diamond	75.8%	#109 of 229	Epoch AI run
USAMO 2025	4.76%	#5 of 10	Official leaderboard
IneqMath	3.50%	#18 of 34	Official leaderboard
GSM8K Platinum Initial 268	98.9%	#11 of 19	Official leaderboard
MATH-500	89.8%	#23 of 38	Official leaderboard
PHYBench	21.2%	#17 of 28	Official leaderboard
OTIS Mock AIME 2024-2025	55.6%	#132 of 212	Epoch AI run
SciArena	989	#22 of 34	Official leaderboard
SimpleBench	36.1%	#78 of 106	Official leaderboard
SnorkelSequences	39.2%	#20 of 25	Official leaderboard
FrontierMath	3.79%	#46 of 55	Official leaderboard
FrontierMath Tier 4	0%	#48 of 55	Epoch AI run
ARC-AGI	5.50%	#88 of 95	Official leaderboard
ARC-AGI-2	0%	#94 of 98	Official leaderboard

Coding

Benchmark	Score	Rank	Source
BigCodeBench-Hard	33.1%	#7 of 96	Official leaderboard
Aider Polyglot	53.3%	#19 of 46	Official leaderboard
WeirdML	37.2%	#89 of 112	Official leaderboard
DesignArena Website	1,052	#124 of 151	Official leaderboard
DesignArena Code	1,043	#120 of 143	Official leaderboard
DesignArena UI Component	1,028	#117 of 139	Official leaderboard
DesignArena 3D Design	1,032	#116 of 136	Official leaderboard
DesignArena Game Dev	1,016	#123 of 143	Official leaderboard
DesignArena Data Visualization	1,028	#111 of 118	Official leaderboard
DesignArena SVG	973	#98 of 103	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
Elimination Game	5	#15 of 56	Official leaderboard
TextQuests	15.4	#8 of 28	Official leaderboard
Berkeley Function Calling Leaderboard (BFCL)	63.3%	#21 of 67	Official leaderboard
BALROG	29.5%	#15 of 33	Official leaderboard
LisanBench	739.0	#62 of 133	Official leaderboard
LLM Chess	-137	#86 of 123	Official leaderboard
Step Race	1	#43 of 59	Official leaderboard
SnorkelWordle	71%	#14 of 19	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
TaxEval v2	75.9%	#10 of 129	Official leaderboard
SnorkelUnderwrite	78%	#6 of 27	Official leaderboard
HealthBench	54.3%	#2 of 8	Official leaderboard
HealthBench Hard	22.6%	#2 of 8	Official leaderboard
xbench-Marketing	46.5	#3 of 7	Official leaderboard
LegalBench	82.6%	#67 of 138	Official leaderboard
CorpFin v2	59.7%	#72 of 119	Official leaderboard
Finance Reasoning (Snorkel)	41.8%	#13 of 21	Official leaderboard
FinQA (Snorkel OpenEnv environment)	65.9%	#14 of 22	Official leaderboard
SnorkelFinance	65.9%	#14 of 22	Official leaderboard
Vals Finance Agent v1	24.1%	#30 of 47	Official leaderboard
xbench-Recruitment	47.1	#7 of 8	Official leaderboard
Agents' Last Exam	4.61%	#31 of 31	Official leaderboard

Vision, audio and robotics

Benchmark	Score		Rank	Source
MMMU	78%		#12 of 136	Official leaderboard
EgoSchema	74.5%		#2 of 12	Lab model card

Safety and security

Benchmark	Score	Rank	Source
BaxBench	33.9%	#13 of 34	Official leaderboard
Lakera B3 (Backbone Breaker Benchmark)	53%	#21 of 35	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	49.3%	#45 of 71	Official leaderboard
AIR-Bench 2024	51.3%	#65 of 82	Official leaderboard
Molecular Biology Capabilities Test	41%	#72 of 90	Official leaderboard
SimpleSafetyTests	96.8%	#60 of 75	Official leaderboard
HELM Safety	85.5	#63 of 75	Official leaderboard
KORA	18%	#36 of 42	Official leaderboard
Anthropic Red Team	95.5%	#69 of 75	Official leaderboard
Human Pathogen Capabilities Test	33%	#80 of 86	Official leaderboard

Other

Benchmark	Score		Rank	Source
HELM Capabilities	72.7		#15 of 60	Official leaderboard

Its lab's numbers put it at 142.0, 4.5 points above its ECI

The Self-Reported Index fits one score to the 8 capability results xAI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 139.2 to 144.8. How we compute it.

AIME 2025 93.3
AIME 2024 93.3
GPQA Diamond 84.6
LOFT 83.3
MMLU-Pro 79.9
MMMU 78
EgoSchema 74.5
SimpleQA 43.6

Its lab's numbers next to 4 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap	What differs
GPQA Diamond	84.6% GPQA (Graduate-Level Google-Proof Q&A (Diamond)) (cons@64; pass@1 = 80.2)	75.8% Epoch AI run	+8.8 pts
LMArena (Elo)	1,402 Chatbot Arena Score (early Grok 3 'chocolate')	1,411 Leaderboard	−9	The lab quotes the early chocolate model snapshot, while the board rates the later Grok 3 Preview release with style controls.
MMLU-Pro	79.9% MMLU-pro [Grok 3 Beta, non-reasoning table]	79.9% Leaderboard	Same
MMMU	78% MMMU (Grok 3 Think; pass@1 = 76)	78% Leaderboard	Same

xAI's models over time

xAI's models in color, against every other model on the Epoch Capabilities Index, by release date.

Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Grok 3 Beta — The Age of Reasoning Agents Feb 19, 2025