Grok 4.7 ranks first on 2 of the 261 benchmarks with five or more models scored.
Closed weights Flagship
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.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Longform Creative Writing (EQ-Bench) | 83.0 | #5 of 119 | Official leaderboard | |
| Creative Writing v3 | 79.8 | #13 of 120 | Official leaderboard | |
| BullshitBench V2 | 75% | #16 of 129 | Official leaderboard | |
| Opening Lines Arena | 1,596 | #3 of 15 | Official leaderboard | |
| SimpleQA Verified | 59.4% | #12 of 56 | Official leaderboard | |
| LMArena (Elo) | 1,442 | #74 of 335 | Official leaderboard | |
| HANDBOOK.md | 44% | #4 of 16 | Official leaderboard | |
| Corp Speak Arena | 1,565 | #6 of 22 | Official leaderboard | |
| AA-LCR Extended | 68.2% | #19 of 49 | Official leaderboard | |
| Lech Mazur Creative Writing | 0.35 | #22 of 50 | Official leaderboard | |
| FACTS Grounding v2 | 68.6% | #11 of 25 | Official leaderboard | |
| Extended NYT Connections | 76.8% | #49 of 103 | Official leaderboard | |
| ComplexConstraints | 60.8% | #8 of 15 | Official leaderboard | |
| FACTS Parametric V2 | 54.3% | #15 of 27 | Official leaderboard | |
| FACTS Benchmark Suite | 62.3% | #14 of 23 | Official leaderboard | |
| StudyArena Text | 1,222 | #53 of 85 | Official leaderboard | |
| FACTS Multimodal | 40.9% | #16 of 23 | Official leaderboard | |
| Joke Arena | 1,478 | #19 of 25 | Official leaderboard | |
| ELI5 Arena | 1,371 | #17 of 18 | Official leaderboard | |
| HLE-Diamond | 25% | #19 of 20 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DTBench | 96% | #19 of 180 | Official leaderboard | |
| GPQA Diamond | 92.7% | #27 of 229 | Epoch AI run | |
| OTIS Mock AIME 2024-2025 | 98.1% | #34 of 212 | Epoch AI run | |
| LatchBio Capabilities v1.0 | 47.4% | #9 of 27 | Official leaderboard | |
| ARC-AGI-2 | 61.4% | #33 of 98 | Official leaderboard | |
| ARC-AGI | 90.2% | #35 of 95 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 58.5% | #10 of 26 | Official leaderboard | |
| Terminal-Bench-Science | 14.3% | #8 of 17 | Official leaderboard | |
| FrontierPhysics | 4.20% | #12 of 22 | Official leaderboard | |
| Riemann-bench | 57.6% | #11 of 20 | Official leaderboard | |
| BioMysteryBench | 69.3% | #17 of 30 | Official leaderboard | |
| MysteryMechanism | 25.2% | #18 of 31 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 53% | #51 of 86 | Official leaderboard | |
| ProofBench v1.1 | 26% | #39 of 52 | Official leaderboard | |
| BixBench | 88.4% | #3 of 4 | Lab model card | |
| FrontierMath Tier 4 v2 | 17.1% | #52 of 68 | Official leaderboard | |
| CritPt | 50.6% | #9 of 11 | Official leaderboard | |
| CritPt v2 | 50.6% | #9 of 11 | Official leaderboard | |
| Human Behavior Bench | 75.2 | #10 of 11 | Official leaderboard | |
| FrontierMath Erdős | 0% | #15 of 15 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| WebDev Arena | 1,638 | #12 of 105 | arena.ai, via Epoch AI | |
| Vibe Code Bench | 86.2% | #15 of 120 | Official leaderboard | |
| DAXBench v2 | 91.2% | #6 of 47 | Official leaderboard | |
| APEX-SWE | 53.4% | #10 of 65 | Official leaderboard | |
| DesignArena React Native | 1,250 | #9 of 56 | Official leaderboard | |
| Code Arena | 1,641 | #21 of 128 | Official leaderboard | |
| DesignArena Game Dev | 1,307 | #26 of 143 | Official leaderboard | |
| SWE-Together | 65.1% | #4 of 22 | Official leaderboard | |
| RSI Arena | 1,038 | #2 of 10 | Official leaderboard | |
| DesignArena 3D Design | 1,259 | #32 of 136 | Official leaderboard | |
| DesignArena Agentic Overall | 1,243 | #18 of 74 | Official leaderboard | |
| Code Migration | 44.8% | #19 of 78 | Official leaderboard | |
| ALE-Bench | 1,823 | #33 of 126 | Official leaderboard | |
| Doc2Harness | 43.5% | #5 of 19 | Official leaderboard | |
| DesignArena Code | 1,267 | #38 of 143 | Official leaderboard | |
| DesignArena UI Component | 1,274 | #37 of 139 | Official leaderboard | |
| DeepSWE | 71% | #18 of 56 | Lab model card | |
| FrontierCode 2.0 | 28.9% | #11 of 32 | Official leaderboard | |
| Image-to-WebDev Arena | 1,589 | #19 of 55 | Official leaderboard | |
| OpenDesign Arena | 82.9 | #8 of 21 | Official leaderboard | |
| DesignArena Website | 1,245 | #59 of 151 | Official leaderboard | |
| MirrorCode | 54.4% | #16 of 39 | Epoch AI run | |
| IOI v2 (Vals) | 57.7% | #19 of 45 | Official leaderboard | |
| SWE Atlas v1.1 | 62.1% | #12 of 28 | Official leaderboard | |
| Next.js Agent Evals | 96% | #6 of 14 | Official leaderboard | |
| Code Migration v1.1 | 53.2% | #13 of 30 | Official leaderboard | |
| SWE-Marathon | 46% | #5 of 11 | Lab model card | |
| DesignArena Data Visualization | 1,224 | #54 of 118 | Official leaderboard | |
| Chart Arena | 1,509 | #7 of 15 | Official leaderboard | |
| DesignArena Image to HTML | 1,209 | #24 of 51 | Official leaderboard | |
| SWE Atlas Codebase QnA Extended | 15.3% | #19 of 38 | Official leaderboard | |
| AutoResearchExam | 0.52 | #6 of 12 | Official leaderboard | |
| Senior SWE-Bench | 27.4% | #12 of 23 | Official leaderboard | |
| SWE-bench Verified Extended | 55.3% | #23 of 44 | Official leaderboard | |
| CursorBench | 46.3% | #11 of 21 | cursor.com, via Epoch AI | |
| DAYJOB: Data Science | 6.70% | #8 of 15 | Official leaderboard | |
| SWE-Marathon v1.1 | 45.6% | #11 of 20 | Official leaderboard | |
| RSI-Exam | 38.4% | #10 of 18 | Official leaderboard | |
| ViBench 2.0 | 1.50% | #5 of 9 | Official leaderboard | |
| FrontierSWE | 29.5% | #14 of 25 | Official leaderboard | |
| SWE-bench CLI | 6.10% | #9 of 16 | Official leaderboard | |
| CursorBench 4.0 | 46.3% | #12 of 21 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 31.9% | #10 of 16 | Official leaderboard | |
| WeirdML v3 | 8.45% | #17 of 27 | Official leaderboard | |
| ReactBench | 32.3% | #18 of 28 | Official leaderboard | |
| Vals RSI Index | 0.20 | #18 of 27 | Official leaderboard | |
| Real-SWE 1.1 | 28% | #10 of 15 | Official leaderboard | |
| Brokk Power Ranking | 5.40 | #14 of 19 | Official leaderboard | |
| Landing Page Arena | 1,411 | #11 of 14 | Official leaderboard | |
| Web Clone Bench | 2.80% | #8 of 10 | Official leaderboard | |
| OWE-Bench | 7.00% | #12 of 14 | Official leaderboard | |
| DeepRSI | 685 | #7 of 8 | Official leaderboard | |
| sudo L7 | 12.2% | #16 of 18 | Official leaderboard | |
| Floorplan Arena | 1,270 | #10 of 11 | Official leaderboard | |
| FrontierMemory | 37.8% | #7 of 7 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Hermes Bench | 74.4% | #2 of 17 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 534 | #5 of 39 | Official leaderboard | |
| Game Arena Nine Men's Morris | 896 | #8 of 39 | Official leaderboard | |
| LLM Chess | 822 | #30 of 123 | Official leaderboard | |
| Game Arena Lines of Action | 570 | #10 of 39 | Official leaderboard | |
| RuneBench | 5.66 | #21 of 76 | Official leaderboard | |
| DuelLab GameBench 2 | 72.3 | #17 of 57 | Official leaderboard | |
| Agent Arena | 2.56% | #17 of 56 | Official leaderboard | |
| BrowseComp Extended | 39.5% | #13 of 42 | Official leaderboard | |
| Game Arena Four in a Row | 474 | #15 of 43 | Official leaderboard | |
| Terminal-Bench 4.1 | 38.8% | #15 of 43 | Official leaderboard | |
| ARC-AGI-3 | 10.1% | #10 of 27 | Official leaderboard | |
| Terminal-Bench 4.0 | 37.6% | #16 of 40 | Official leaderboard | |
| Kaggle Game Arena | 336 | #17 of 42 | Official leaderboard | |
| Chess Puzzles | 41% | #38 of 93 | Epoch AI run | |
| Mystery Game Puzzles | 29% | #36 of 84 | Epoch AI run | |
| Game Arena Clobber | 330 | #17 of 39 | Official leaderboard | |
| Game Arena Coin Game | 382 | #17 of 39 | Official leaderboard | |
| Collaborative Intelligence Benchmark | 55.6% | #8 of 18 | Official leaderboard | |
| Game Arena Chess | 842 | #19 of 42 | Official leaderboard | |
| Game Arena Dots and Boxes | 917 | #19 of 40 | Official leaderboard | |
| Game Arena Go | 1,004 | #19 of 39 | Official leaderboard | |
| Game Arena Word Association | 279 | #20 of 40 | Official leaderboard | |
| BU Bench V2.1 | 73.3 | #8 of 16 | Official leaderboard | |
| WitnessBench | 14.8% | #6 of 12 | Official leaderboard | |
| Game Arena Dark Hex | 262 | #21 of 41 | Official leaderboard | |
| Game Arena Word Art | 491 | #21 of 39 | Official leaderboard | |
| Game Arena Reversi | 453 | #23 of 39 | Official leaderboard | |
| o11y-bench | 76.6% | #13 of 22 | Official leaderboard | |
| Game Arena Checkers | 888 | #26 of 41 | Official leaderboard | |
| Game Arena Bargaining | 163 | #26 of 39 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 11.8% | #42 of 44 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnorkelUnderwrite 2.0 | 30.5% | #1 of 16 | Official leaderboard | |
| MedScribe | 89.4% | #7 of 97 | Official leaderboard | |
| DAYJOB: Healthcare | 27.6% | #5 of 37 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 68.2% | #7 of 51 | Official leaderboard | |
| Vending-Bench 2 | $10,537 | #10 of 67 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 12.5% | #13 of 80 | Official leaderboard | |
| Legal Research Bench | 47.1% | #17 of 93 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 18.3% | #7 of 38 | Official leaderboard | |
| Vals Index v2 | 60.2% | #12 of 65 | Official leaderboard | |
| SnorkelFinance 2.0 | 20.2% | #3 of 16 | Official leaderboard | |
| Tax Agent Bench | 35.8% | #17 of 84 | Official leaderboard | |
| Tuesday Work Index | 64.7 | #9 of 44 | Official leaderboard | |
| DAYJOB: Finance | 26.5% | #8 of 39 | Official leaderboard | |
| MedCode | 49.5% | #22 of 96 | Official leaderboard | |
| LegalBench | 84.4% | #34 of 138 | Official leaderboard | |
| Vending-Bench Arena | $7,930 | #8 of 32 | Official leaderboard | |
| DAYJOB: Law | 27.6% | #4 of 16 | Official leaderboard | |
| Ivo-micro1 ContractBench | 58.3% | #2 of 8 | Official leaderboard | |
| FoodTruck Bench | $42,561 | #15 of 59 | Official leaderboard | |
| Excel Modeling Benchmark | 67% | #20 of 76 | Official leaderboard | |
| Vals Legal Index | 72.7% | #10 of 38 | Official leaderboard | |
| EEBench | 64% | #8 of 30 | Official leaderboard | |
| DAYJOB: Tax | 5.00% | #4 of 15 | Official leaderboard | |
| YC-Bench | $1,666,389 | #16 of 59 | Official leaderboard | |
| SnorkelManufacturing | 12.1% | #5 of 16 | Official leaderboard | |
| PolicyBench | 87.6% | #16 of 50 | Official leaderboard | |
| Short Form Bench | $2,578 | #3 of 9 | Official leaderboard | |
| Vals Index v2.1 | 55% | #17 of 49 | Official leaderboard | |
| GDP.xlsx | 22.3% | #7 of 20 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 71.5% | #15 of 42 | Official leaderboard | |
| WorkplaceAgents | 15.2% | #5 of 14 | Official leaderboard | |
| Vals Finance Index | 58.8% | #15 of 40 | Official leaderboard | |
| Physical-Engineering's Last Exam | 0.30 | #7 of 18 | Official leaderboard | |
| APEX-Agents 1.1 | 68.7% | #26 of 65 | Official leaderboard | |
| Immigration Bench | 68% | #10 of 25 | Official leaderboard | |
| DAYJOB: Accounting | 20% | #7 of 17 | Official leaderboard | |
| Public Benefits Bench | 65.6% | #21 of 50 | Official leaderboard | |
| APEX-Agents | 54.6% | #26 of 61 | Collected by Epoch AI, source not named | |
| APEX-1 | 65.4% | #31 of 71 | Official leaderboard | |
| Vals Finance Agent v2 | 52.3% | #36 of 80 | Official leaderboard | |
| PRBench | 47.6% | #22 of 48 | Official leaderboard | |
| TaxCalcBench v2 | 48% | #15 of 32 | Official leaderboard | |
| Legora BAR | 102.0 | #8 of 17 | Official leaderboard | |
| SnorkelLegal | 32.8% | #8 of 16 | Official leaderboard | |
| CADGenBench | 44.4% | #1 of 2 | Lab model card | |
| Vals Index v2.2 | 58.6% | #19 of 37 | Official leaderboard | |
| Parametric CAD Bench | 34.8% | #6 of 11 | Official leaderboard | |
| Digital Work Index | 19.3 | #9 of 16 | Official leaderboard | |
| DeepFinance | 34.1% | #13 of 23 | Official leaderboard | |
| RCM-Bench-1 | 88.2% | #12 of 20 | Official leaderboard | |
| HW-Bench | 20% | #9 of 15 | Official leaderboard | |
| LegalReasonBench | 40.4% | #6 of 10 | Official leaderboard | |
| MedXpertQA MM Extended | 52% | #25 of 41 | Official leaderboard | |
| Holdings-IndividualTaxBench | 72.3% | #8 of 13 | Official leaderboard | |
| RealmPathology | 73.1% | #8 of 13 | Official leaderboard | |
| DesignArena PDF | 1,168 | #23 of 37 | Official leaderboard | |
| AA-Briefcase v1.1 | 1,657 | #5 of 8 | Lab model card | |
| Vals Finance Agent v2.1 | 60.6% | #26 of 40 | Official leaderboard | |
| Thrive Holdings helpdesk ticket resolution benchmark | 46.1% | #10 of 15 | Official leaderboard | |
| SnorkelRevOps | 11.3% | #11 of 16 | Official leaderboard | |
| APEX-Accounting | 55.3% | #35 of 49 | Official leaderboard | |
| Clipping Bench | $384 | #14 of 19 | Official leaderboard | |
| GDPx | 3.87% | #15 of 20 | Official leaderboard | |
| CAD Arena | 0.26 | #9 of 12 | Official leaderboard | |
| HealthBench Professional | 56.7% | #21 of 26 | Lab model card | |
| RWE-bench | 0% | #13 of 15 | Official leaderboard | |
| MissionReport | 59.5% | #9 of 10 | Official leaderboard | |
| SheetBench | 14.3% | #8 of 8 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,829 | #15 of 75 | Official leaderboard | |
| BVB | 70.1 | #12 of 45 | Official leaderboard | |
| Vision Arena | 1,249 | #45 of 136 | Official leaderboard | |
| GDP.pdf | 48.8% | #8 of 19 | Official leaderboard | |
| Blueprint-Bench 2 | 32.5% | #18 of 38 | Official leaderboard | |
| CharXiv Extended | 78.2% | #26 of 38 | Official leaderboard | |
| DesignArena Code to Video | 1,110 | #21 of 29 | Official leaderboard | |
| GeoBench | 61% | #35 of 48 | Official leaderboard | |
| Chartography | 14.7% | #34 of 43 | Official leaderboard | |
| SplatBench | 2.10 | #7 of 8 | Official leaderboard | |
| DataCenterBench Tier 1 | 13.3% | #8 of 9 | Official leaderboard | |
| 1H-VideoQA | 59.4% | #19 of 20 | Official leaderboard | |
| Furniture Assembly Benchmark | 20.8% | #34 of 35 | Official leaderboard | |
| Dual-Arm Sim | 6.30 | #11 of 11 | Official leaderboard | |
| Robo-bench | 0.21% | #7 of 7 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BioSecBench-Refusal | 62.4% | #1 of 22 | Official leaderboard | |
| ABC Bench Fragment Design | 100% | #2 of 27 | Official leaderboard | |
| Virology Capabilities Test (VCT) | 63% | #3 of 31 | Lab model card | |
| World-Class Bio | 57.6% | #12 of 74 | Official leaderboard | |
| dfbench | 59% | #4 of 18 | Official leaderboard | |
| BioTIER Refuse | 80.3% | #16 of 69 | Official leaderboard | |
| CWE-bench | 68% | #3 of 12 | Official leaderboard | |
| VariantBench | 45.5% | #7 of 27 | Official leaderboard | |
| MetagenomicsBench | 48.7% | #7 of 26 | Official leaderboard | |
| SAFE-Teen Health advice | 39% | #5 of 17 | Official leaderboard | |
| TxBench-Antibody-Discovery | 52% | #9 of 30 | Official leaderboard | |
| ABC Bench Liquid-Handling Robot | 98% | #11 of 34 | Official leaderboard | |
| BioSecBench-Function | 43.3% | #9 of 27 | Official leaderboard | |
| Standard jailbreaks (xAI) | 0.01% | #1 of 3 | Lab model card | |
| SAFE-Teen Relationship boundaries | 48% | #6 of 17 | Official leaderboard | |
| BioSecBench-Surveillance | 48% | #10 of 27 | Official leaderboard | |
| Biopharma Bench | 56.3% | #5 of 13 | Official leaderboard | |
| TxBench-Oligo-Discovery | 41.9% | #12 of 30 | Official leaderboard | |
| Human Pathogen Capabilities Test | 53.7% | #39 of 86 | Official leaderboard | |
| OncologyBench | 18.7% | #14 of 30 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 54.1% | #35 of 71 | Official leaderboard | |
| CyberGym | 80.3% | #17 of 34 | Lab model card | |
| xAI sycophancy eval | 0.03% | #2 of 4 | Lab model card | |
| BioUseBench | 91.4% | #1 of 2 | Lab model card | |
| Long-horizon jailbreaks (xAI) | 0.65% | #1 of 2 | Lab model card | |
| CyberBench v1.1 (Vals) | 69.5% | #23 of 44 | Official leaderboard | |
| VCT-v2 | 42% | #14 of 25 | Official leaderboard | |
| SRE-Bench | 3.82% | #20 of 35 | Official leaderboard | |
| TxBench-Preclinical-Pharmacology | 54.7% | #18 of 30 | Official leaderboard | |
| Molecular Biology Capabilities Test | 46.1% | #57 of 90 | Official leaderboard | |
| CWE-bench 2 | 54.7% | #9 of 14 | Official leaderboard | |
| SAFE-Teen Urgent safety | 44% | #11 of 17 | Official leaderboard | |
| BioTIER Permit | 99.2% | #45 of 69 | Official leaderboard | |
| CyberOffense | 54.4% | #4 of 6 | Official leaderboard | |
| Autointent-Bio | 100% | #2 of 3 | Lab model card | |
| Autointent-Chem | 99.9% | #2 of 3 | Lab model card | |
| LAB-Bench | 76.8% | #2 of 3 | Lab model card | |
| EpiBench | 20.8% | #19 of 27 | Official leaderboard | |
| SpatialBench-Long | 24.4% | #20 of 27 | Official leaderboard | |
| scBench-Long | 11.4% | #22 of 27 | Official leaderboard | |
| FrontierBioML | 24.0 | #9 of 11 | Official leaderboard | |
| DeepsecBench | 10.9 | #24 of 29 | Official leaderboard | |
| WellnessBench | 48% | #10 of 12 | Official leaderboard | |
| VIALS 2 | 18.2% | #12 of 14 | Official leaderboard | |
| General refusals (xAI) | 1.10% | #7 of 8 | Lab model card | |
| CheatBench | 78% | #9 of 10 | Official leaderboard | |
| BioSimBench | 0.12 | #11 of 12 | Official leaderboard | |
| CVE-Bench | 36.6% | #13 of 14 | Lab model card | |
| DistressBench | 52.4% | #41 of 42 | Official leaderboard | |
| ProtocolQA Open-Ended (xAI) | 70.4% | #4 of 4 | Lab model card | |
| Child safety (xAI) | 0% | #3 of 3 | Lab model card | |
| FORTRESS-RN | 97.9% | #3 of 3 | Lab model card | |
| HackerBench | 4.02% | #3 of 3 | Lab model card | |
| Self-harm refusals (xAI) | 1.05% | #3 of 3 | Lab model card | |
| Biosecurity VCT | 41.5% | #2 of 2 | Lab model card | |
| CathedralBench | 29% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 63.1 | #29 of 151 | Official leaderboard | |
| ACCoRD | 74.0 | #31 of 152 | Official leaderboard | |
| LMCA | 49.4% | #31 of 147 | Official leaderboard | |
| SAGE v1.1 (Vals) | 88.7% | #7 of 29 | Official leaderboard | |
| Hermes Index | 39.3 | #6 of 17 | Official leaderboard | |
| SuperCLUE | 49.7 | #6 of 16 | Official leaderboard | |
| Political Science LLM Benchmark | 66.6% | #20 of 42 | Official leaderboard | |
| LiveBench | 67.9% | #15 of 31 | Official leaderboard | |
| Vals Science Index | 34.8% | #16 of 25 | Official leaderboard | |
| SAGE (Vals) | 40.8% | #58 of 81 | Official leaderboard | |
| Xbench | -11.11 | #15 of 17 | Official leaderboard |
The Self-Reported Index fits one score to the 6 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 152.2 to 162.4. How we compute it.
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 |
|---|---|---|---|
| Harvey LAB (Legal Agent Benchmark) | 19.6% Harvey Legal Agent Benchmark (Legal work) | 12.5% Leaderboard | +7.1 pts |
| LatchBio Capabilities v1.0 | 44.5% LatchBio Capabilities v1.0, overall score (mean of 11 capability benchmarks), xhigh (5.2) | 47.4% Leaderboard, Pi | −2.9 pts |
| EEBench | 66% EEBench V1 core corpus, reward, xhigh, Grok Build harness, run by Atopile (4.1) | 64% Leaderboard, Grok CLI | +2.0 pts |
| FrontierSWE | 29% FrontierSWE V2, mean@5 partial-credit score, xhigh, run by Proximal Labs (2.4) | 29.5% Leaderboard, mean@5, proximus | −0.5 pts |
| Terminal-Bench 4.0 | 38% Terminal-Bench 4.0, task success rate, xhigh, Grok Build harness, run by Harbor (2.3) | 37.6% Leaderboard, xhigh, Grok Build | +0.4 pts |
| BioSecBench-Surveillance | 48% BioSecBench Surveillance, endpoint pass rate, xhigh, Grok Build harness (7.1 table) | 48% Leaderboard, Pi | Same |
| BioSecBench-Function | 43.3% BioSecBench Function, endpoint pass rate, xhigh, Grok Build harness (7.1 table) | 43.3% Leaderboard, Pi | Same |
| BioSecBench-Refusal | 62.4% BioSecBench Refusal, xhigh, Grok Build harness (7.1 table) | 62.4% Leaderboard, Pi | Same |
| CursorBench 4.0 | 46.3% CursorBench 4.0 (Software engineering) | 46.3% Leaderboard, xhigh, Cursor | Same |
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)
The benchmarks the lab chose to headline, with the scores it gave.
Every result in the longer documents the lab published with it. 29 of them aren't in the launch post.