Rankings / xAI
Grok 3
released 2025-04-09
BenchAtlas Index
as of 2026-07-14
base configuration
rank #292
6 families · 3 categories · medium
base configuration
rank #343
10 families · 5 categories · high
Benchmark evidence
22 results
agentic coding
| τ²-Bench | 48.8 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 42.5 % | independent | Artificial Analysis | |
| SciCode | 36.8 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 18.4 points | independent | Artificial Analysis | |
| AA Math Index | 58.0 points | independent | Artificial Analysis | |
| ECI | 139.0 points | independent | Epoch AI Benchmarking Hub |
knowledge science
| GPQA Diamond implementation=vals-ai | 74.2 % | independent | Vals AI | |
| GPQA Diamond implementation=artificial-analysis | 69.3 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 5.1 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=vals-ai | 79.9 % | independent | Vals AI | |
| MMLU-Pro implementation=artificial-analysis | 79.9 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | 54.7 % | independent | Artificial Analysis | |
| IFBench | 46.9 % | independent | Artificial Analysis |
professional
| CorpFin | 59.7 % | independent | Vals AI | |
| LegalBench | 82.6 % | independent | Vals AI | |
| MedQA | 83.8 % | independent | Vals AI | |
| TaxEval | 75.9 % | independent | Vals AI |
reasoning math
| AIME implementation=vals-ai | 58.8 % | independent | Vals AI | |
| AIME year=2025 · implementation=artificial-analysis | 58.0 % | independent | Artificial Analysis | |
| AIME implementation=artificial-analysis | 33.0 % | independent | Artificial Analysis | |
| MATH-500 implementation=vals-ai | 89.8 % | independent | Vals AI | |
| MATH-500 implementation=artificial-analysis | 87.0 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model Artificial Analysis harness + Grok 3 | Terminal-Bench Hard | 11.4 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.