GLM-5.3 ranks first on 4 of the 269 benchmarks with five or more models scored.
Open 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 | |
|---|---|---|---|---|
| Creative Writing v3 | 82.2 | #2 of 120 | Official leaderboard | |
| AA-LCR Extended | 76.8% | #2 of 49 | Official leaderboard | |
| Corp Speak Arena | 1,731 | #1 of 22 | Official leaderboard | |
| LMArena (Elo) | 1,483 | #23 of 335 | Official leaderboard | |
| Humanity's Last Exam | 62.5% | #11 of 154 | Lab model card | |
| Longform Creative Writing (EQ-Bench) | 81.8 | #10 of 119 | Official leaderboard | |
| Lech Mazur Creative Writing | 2.95 | #5 of 50 | Official leaderboard | |
| BullshitBench V2 | 72% | #20 of 129 | Official leaderboard | |
| Antidote | 1,067 | #8 of 36 | Official leaderboard | |
| ELI5 Arena | 1,596 | #4 of 18 | Official leaderboard | |
| StudyArena Text | 1,257 | #21 of 85 | Official leaderboard | |
| BullshitBench | 50.9% | #28 of 107 | Official leaderboard | |
| Hemingway-bench | 1,053 | #11 of 42 | Official leaderboard | |
| OpenAI MRCR v2 | 57.8% | #13 of 47 | Official leaderboard | |
| Joke Arena | 1,540 | #7 of 25 | Official leaderboard | |
| Judgemark v4 | 72.6 | #12 of 40 | Official leaderboard | |
| DesignArena ASCII Art | 1,210 | #26 of 77 | Official leaderboard | |
| AgentIF (2026-10 setting) | 64.2% | #9 of 24 | Official leaderboard | |
| Last Translation Benchmark | 8.56 | #18 of 36 | Official leaderboard | |
| Extended NYT Connections | 74.2% | #55 of 103 | Official leaderboard | |
| HANDBOOK.md | 29.9% | #9 of 16 | Official leaderboard | |
| SimpleQA Verified | 41% | #54 of 94 | Epoch AI run | |
| Opening Lines Arena | 1,455 | #9 of 15 | Official leaderboard | |
| ComplexConstraints | 54.4% | #10 of 15 | Official leaderboard | |
| FACTS Parametric V2 | 40.3% | #21 of 27 | Official leaderboard | |
| HLE-Diamond | 25.1% | #18 of 20 | Official leaderboard | |
| Continual Learning Exam | 24% | #11 of 11 | Official leaderboard | |
| Kimi Hallu Bench | 14.8% | #6 of 6 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 90.9% | #40 of 229 | Epoch AI run | |
| ASI-Bench | 41.6 | #3 of 11 | Official leaderboard | |
| SimpleBench | 66.2% | #29 of 106 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 91.1% | #68 of 212 | Epoch AI run | |
| FrontierMath Tiers 1-3 v2 | 68.8% | #30 of 86 | Official leaderboard | |
| DTBench | 87.7% | #63 of 180 | Official leaderboard | |
| Prophet Arena | 85.8 | #10 of 24 | Official leaderboard | |
| FrontierPhysics | 4.20% | #11 of 22 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 29.3% | #36 of 68 | Official leaderboard | |
| NatureBench | 10% | #11 of 18 | Official leaderboard | |
| ProofBench v1.1 | 49% | #32 of 52 | Official leaderboard | |
| MysteryMechanism | 23% | #21 of 31 | Official leaderboard | |
| PaperBenchX | 7.53% | #7 of 10 | Official leaderboard | |
| Terminal-Bench-Science | 8.10% | #12 of 17 | Official leaderboard | |
| Apodex Discovery | 2.45 | #6 of 8 | Official leaderboard | |
| Into the Unknown | 0% | #7 of 9 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 48.7% | #22 of 26 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 33.6% | #23 of 27 | Official leaderboard | |
| xbench-ScienceQA | 65.8% | #26 of 29 | Official leaderboard | |
| Riemann-bench | 20.8% | #19 of 20 | Official leaderboard | |
| BTF-3 | 0.15 | #8 of 8 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FrogsGame post-training task | 33% | #2 of 17 | Official leaderboard | |
| FrontierSWE v1 | 78.1% | #2 of 17 | Official leaderboard | |
| WeirdML | 75.4% | #17 of 112 | Official leaderboard | |
| WebDev Arena | 1,614 | #17 of 105 | arena.ai, via Epoch AI | |
| DesignArena 3D Design | 1,299 | #26 of 136 | Official leaderboard | |
| DesignArena Website | 1,280 | #29 of 151 | Official leaderboard | |
| ALE-Bench | 1,937 | #25 of 126 | Official leaderboard | |
| DesignArena Code | 1,285 | #29 of 143 | Official leaderboard | |
| DesignArena Frontend Web App | 1,270 | #11 of 53 | Official leaderboard | |
| DesignArena Game Dev | 1,290 | #30 of 143 | Official leaderboard | |
| Code Arena | 1,622 | #27 of 128 | Official leaderboard | |
| DesignArena SVG | 1,248 | #22 of 103 | Official leaderboard | |
| SWE-bench Verified Extended | 68% | #10 of 44 | Official leaderboard | |
| DesignArena UI Component | 1,284 | #33 of 139 | Official leaderboard | |
| DeepSWE | 69% | #7 of 28 | Official leaderboard | |
| NL2Repo-Bench | 58% | #8 of 30 | Lab model card | |
| Tencent WorkBuddy Bench | 77.4% | #4 of 14 | Official leaderboard | |
| DesignArena Agentic Overall | 1,226 | #22 of 74 | Official leaderboard | |
| TasteVal | 59.2× | #6 of 20 | Official leaderboard | |
| Code Migration | 44.2% | #24 of 78 | Official leaderboard | |
| IOI v2 (Vals) | 68.4% | #14 of 45 | Official leaderboard | |
| Doc2Harness | 42.1% | #6 of 19 | Official leaderboard | |
| Vibe Code Bench | 78.1% | #38 of 120 | Official leaderboard | |
| DesignArena Data Visualization | 1,253 | #40 of 118 | Official leaderboard | |
| RSI-Exam | 40.3% | #7 of 18 | Official leaderboard | |
| Code Migration v1.1 | 55.1% | #12 of 30 | Official leaderboard | |
| RSI Arena | 1,024 | #4 of 10 | Official leaderboard | |
| Z.ai Code Bench | 34.5% | #2 of 5 | Official leaderboard | |
| APEX-SWE | 45.6% | #27 of 65 | Official leaderboard | |
| FrontierMemory | 68.9% | #3 of 7 | Official leaderboard | |
| PostTrainBench | 39.8% | #3 of 7 | Lab model card | |
| SlopCodeBench | 24% | #10 of 22 | Official leaderboard | |
| DesignArena React Native | 1,186 | #26 of 56 | Official leaderboard | |
| FrontierCS 2 | 22.9% | #9 of 19 | Official leaderboard | |
| Vibe Code Bench 1-100 | 20% | #13 of 26 | Official leaderboard | |
| MacroscopeBench | 77.4% | #8 of 16 | Official leaderboard | |
| FrontierSWE | 30.2% | #13 of 25 | Official leaderboard | |
| PostTrainBench 1.2 | 34.5% | #11 of 21 | Official leaderboard | |
| SWE-Marathon | 42.5% | #6 of 11 | Lab model card | |
| Vals RSI Index | 0.23 | #15 of 27 | Official leaderboard | |
| Data Agent Benchmark | 78.3% | #10 of 18 | Official leaderboard | |
| CursorBench | 42.6% | #12 of 21 | cursor.com, via Epoch AI | |
| Landing Page Arena | 1,477 | #8 of 14 | Official leaderboard | |
| DAXBench v2 | 77.3% | #27 of 47 | Official leaderboard | |
| SWE Atlas v1.1 | 58.1% | #17 of 28 | Official leaderboard | |
| SWE-Bench Pro V2 | 95.6% | #11 of 18 | Official leaderboard | |
| SWE-Bench Pro V2 Hard | 84.3% | #11 of 18 | Official leaderboard | |
| CursorBench 4.0 | 42.6% | #13 of 21 | Official leaderboard | |
| LibraryDesignBench | 41.9 | #5 of 8 | Official leaderboard | |
| SWE-Marathon v1.1 | 42.5% | #13 of 20 | Official leaderboard | |
| SWE-Clar | 54% | #6 of 9 | Official leaderboard | |
| ReactBench | 30.8% | #19 of 28 | Official leaderboard | |
| FrontierCode 2.0 | 20.1% | #22 of 32 | Official leaderboard | |
| DAYJOB: Data Science | 4.00% | #11 of 15 | Official leaderboard | |
| Real-SWE 1.1 | 26.4% | #11 of 15 | Official leaderboard | |
| ViBench 2.0 | 0% | #7 of 9 | Official leaderboard | |
| OWE-Bench | 11.8% | #11 of 14 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 24.9% | #13 of 16 | Official leaderboard | |
| SWE-bench CLI | 3.50% | #13 of 16 | Official leaderboard | |
| RSI Bench | 32.3 | #7 of 8 | Official leaderboard | |
| Brokk Power Ranking | 3.00 | #17 of 19 | Official leaderboard | |
| sudo L7 | 9.40% | #18 of 18 | Official leaderboard | |
| ProgramBench almost solved | 19% | #8 of 8 | Lab model card | |
| Kimi Code Bench 3.0 Long-horizon | 45.5% | #6 of 6 | Official leaderboard | |
| Legacy-Bench | 23% | #6 of 6 | Official leaderboard | |
| TestPrism | 11.3% | #6 of 6 | Official leaderboard | |
| StepCodeBench | 40.2% | #5 of 5 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CivBench | 1,634 | #1 of 19 | Official leaderboard | |
| AppWorld Test-Normal | 94% | #1 of 12 | Official leaderboard | |
| AppWorld Test-Challenge | 89.5% | #2 of 12 | Official leaderboard | |
| Terminal-Bench 2.1 | 88.2% | #10 of 56 | Lab model card | |
| MCP Atlas | 84.2% | #8 of 37 | Official leaderboard | |
| DuetBench | 1,032 | #2 of 9 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 39.4% | #11 of 44 | Official leaderboard | |
| Long-Horizon-Terminal-Bench 2.0 | 54.2% | #4 of 15 | Official leaderboard | |
| DuelLab GameBench 2 | 72.3 | #16 of 57 | Official leaderboard | |
| Terminal-Bench 4.0 | 41.8% | #12 of 40 | Official leaderboard | |
| Terminal-Bench 4.1 | 41.2% | #13 of 43 | Official leaderboard | |
| LLM Chess | 605 | #38 of 123 | Official leaderboard | |
| AutomationBench | 48.2% | #16 of 47 | Lab model card | |
| Mystery Game Puzzles | 33% | #30 of 84 | Epoch AI run | |
| Terminal-Bench 3.0 | 32.4% | #4 of 11 | Official leaderboard | |
| BabaIsBench | 96% | #13 of 32 | Official leaderboard | |
| Hermes Bench | 71.8% | #7 of 17 | Official leaderboard | |
| WitnessBench | 16.6% | #5 of 12 | Official leaderboard | |
| RuneBench | 5.00 | #33 of 76 | Official leaderboard | |
| TraceTrial-Bench | 74.6% | #4 of 9 | Official leaderboard | |
| o11y-bench | 81.3% | #10 of 22 | Official leaderboard | |
| Chess Puzzles | 37% | #43 of 93 | Epoch AI run | |
| Agent Arena | 0.99% | #27 of 56 | Official leaderboard | |
| Web Context (Nimble) | 69.8% | #7 of 11 | Official leaderboard | |
| Cresta-Bench | 43.1% | #9 of 14 | Official leaderboard | |
| Toolathlon-Verified | 73% | #15 of 21 | Lab model card | |
| Collaborative Intelligence Benchmark | 45.8% | #13 of 18 | Official leaderboard | |
| RecreationBench | 24.4% | #8 of 10 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| JobBench | 61.4% | #1 of 16 | Official leaderboard | |
| Tax Agent Bench | 43.3% | #9 of 84 | Official leaderboard | |
| MedScribe | 88.8% | #11 of 97 | Official leaderboard | |
| Legal Research Bench | 49% | #11 of 93 | Official leaderboard | |
| SnorkelFinance 2.0 | 21.2% | #2 of 16 | Official leaderboard | |
| SnorkelUnderwrite 2.0 | 30.4% | #2 of 16 | Official leaderboard | |
| GDPval-AA v2 | 1,769 | #4 of 27 | Lab model card | |
| Vals Legal Index | 74.6% | #6 of 38 | Official leaderboard | |
| Public Benefits Bench | 68.5% | #10 of 50 | Official leaderboard | |
| LegalBench | 84.8% | #29 of 138 | Official leaderboard | |
| Vals Finance Index | 61.4% | #10 of 40 | Official leaderboard | |
| SnorkelLegal | 37.8% | #4 of 16 | Official leaderboard | |
| Vending-Bench 2 | $8,164 | #17 of 67 | Official leaderboard | |
| Vals Finance Agent v2 | 55.8% | #22 of 80 | Official leaderboard | |
| Vending-Bench Arena | $7,841 | #9 of 32 | Official leaderboard | |
| DesignArena Python-PPTX Slides | 1,205 | #11 of 36 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 14.6% | #12 of 38 | Official leaderboard | |
| E-Commerce Bench | 3.91 | #6 of 19 | Official leaderboard | |
| Vals Finance Agent v2.1 | 65% | #13 of 40 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 8.33% | #27 of 80 | Official leaderboard | |
| APEX-Agents | 56.6% | #22 of 61 | Collected by Epoch AI, source not named | |
| EnterpriseBench: CoreCraft | 51.8% | #19 of 51 | Official leaderboard | |
| SnorkelManufacturing | 11.9% | #6 of 16 | Official leaderboard | |
| Vals Index v2 | 57% | #25 of 65 | Official leaderboard | |
| Investor's Last Exam | 62.3 | #4 of 10 | Official leaderboard | |
| Vals Index v2.2 | 59.9% | #15 of 37 | Official leaderboard | |
| Vals Index v2.1 | 53.5% | #20 of 49 | Official leaderboard | |
| Legora BAR | 102.0 | #7 of 17 | Official leaderboard | |
| DAYJOB: Finance | 11% | #17 of 39 | Official leaderboard | |
| GDP.xlsx | 21.4% | #9 of 20 | Official leaderboard | |
| $OneMillion-Bench | 31% | #6 of 13 | Official leaderboard | |
| Holdings-IndividualTaxBench | 72.8% | #6 of 13 | Official leaderboard | |
| APEX-Accounting | 59.5% | #23 of 49 | Official leaderboard | |
| Clipping Bench | $644 | #9 of 19 | Official leaderboard | |
| TaxEval v2 | 72.4% | #62 of 129 | Official leaderboard | |
| DAYJOB: Healthcare | 4.40% | #18 of 37 | Official leaderboard | |
| MedCode | 42.9% | #47 of 96 | Official leaderboard | |
| Tuesday Work Index | 50.8 | #22 of 44 | Official leaderboard | |
| FinStepBench DeepResearch | 53.3% | #3 of 6 | Official leaderboard | |
| APEX-Agents 1.1 | 66.2% | #34 of 65 | Official leaderboard | |
| Thrive Holdings helpdesk ticket resolution benchmark | 48.1% | #8 of 15 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 69.1% | #23 of 42 | Official leaderboard | |
| APEX-1 | 63% | #39 of 71 | Official leaderboard | |
| RCM-Bench-1 | 88.3% | #11 of 20 | Official leaderboard | |
| Immigration Bench | 66.7% | #14 of 25 | Official leaderboard | |
| Agents' Last Exam | 28.5% | #15 of 26 | Lab model card | |
| Excel Modeling Benchmark | 56.3% | #46 of 76 | Official leaderboard | |
| PolicyBench | 81.1% | #31 of 50 | Official leaderboard | |
| SnorkelRevOps | 11.8% | #10 of 16 | Official leaderboard | |
| E-Bench (Tencent) | 71.4% | #5 of 8 | Official leaderboard | |
| Ivo-micro1 ContractBench | 55% | #5 of 8 | Official leaderboard | |
| DeepFinance | 27.5% | #16 of 23 | Official leaderboard | |
| Bedside Bench | 86.9% | #12 of 17 | Official leaderboard | |
| DAYJOB: Accounting | 7.60% | #12 of 17 | Official leaderboard | |
| RWE-bench | 0% | #11 of 15 | Official leaderboard | |
| Digital Work Index | 16.4 | #12 of 16 | Official leaderboard | |
| E-Bench-Code (Tencent) | 66.5% | #6 of 8 | Official leaderboard | |
| Quant-Arena | 38.9 | #7 of 9 | Official leaderboard | |
| WorkplaceAgents | 12.8% | #11 of 14 | Official leaderboard | |
| DAYJOB: Tax | 0.40% | #12 of 15 | Official leaderboard | |
| HW-Bench | 17% | #12 of 15 | Official leaderboard | |
| DAYJOB: Law | 6.20% | #13 of 16 | Official leaderboard | |
| ATLAS-Finance | 3.00% | #9 of 11 | Official leaderboard | |
| ATLAS-Finance v1.1.0 | 3.00% | #9 of 11 | Official leaderboard | |
| FinStepBench LiveSearch | 73.3% | #5 of 6 | Official leaderboard | |
| GDPx | 2.69% | #18 of 20 | Official leaderboard | |
| Agents' Last Exam v1.1 | 26.2% | #13 of 14 | Official leaderboard | |
| TaxCalcBench v2 | 0% | #31 of 32 | Official leaderboard | |
| FinStepBench CorporateValuation | 56.1% | #6 of 6 | Official leaderboard | |
| Build-A-Chip-Bench | 0% | #5 of 5 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,740 | #24 of 75 | Official leaderboard | |
| VoxelBench | 1,785 | #25 of 56 | Official leaderboard | |
| Code2Video Benchmark | 1,496 | #10 of 19 | Official leaderboard | |
| GlobeBench | 14.8% | #14 of 19 | Official leaderboard | |
| ArtifactArena | 978 | #16 of 21 | Official leaderboard | |
| Code2Video Arena | 1,449 | #10 of 13 | Official leaderboard | |
| GDP.pdf | 36.6% | #16 of 19 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CyberGym | 97.2% | #2 of 34 | Official leaderboard | |
| SecIT Bench | 81.4% | #2 of 26 | Official leaderboard | |
| ABC Bench Screening Evasion | 67.2% | #5 of 17 | Official leaderboard | |
| ABC Bench Fragment Design | 94.6% | #8 of 27 | Official leaderboard | |
| SRE-Bench | 9.16% | #14 of 35 | Official leaderboard | |
| Human Pathogen Capabilities Test | 55.7% | #35 of 86 | Official leaderboard | |
| CyberBench v1.1 (Vals) | 72.1% | #18 of 44 | Official leaderboard | |
| DeepsecBench | 21.8 | #12 of 29 | Official leaderboard | |
| nocot-bench (NCRI) | 74.5 | #77 of 185 | Official leaderboard | |
| BioTIER Permit | 99.6% | #30 of 69 | Official leaderboard | |
| Molecular Biology Capabilities Test | 49.8% | #40 of 90 | Official leaderboard | |
| DistressBench | 72.5% | #19 of 42 | Official leaderboard | |
| World-Class Bio | 46.6% | #37 of 74 | Official leaderboard | |
| OncologyBench | 17.3% | #15 of 30 | Official leaderboard | |
| ExploitGym | 18.3% | #9 of 18 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 53.2% | #37 of 71 | Official leaderboard | |
| ABC Bench Liquid-Handling Robot | 91.3% | #18 of 34 | Official leaderboard | |
| ExploitBench | 54.4% | #12 of 21 | Lab model card | |
| PCT (Polarized Contrastive Pairs) | 46.8% | #27 of 46 | Official leaderboard | |
| FORTRESS | 28.2% | #44 of 73 | Official leaderboard | |
| BioTIER Refuse | 32.5% | #44 of 69 | Official leaderboard | |
| Last Secure Code Benchmark | 21.3% | #4 of 6 | Official leaderboard | |
| BioSecBench-Refusal | 10.9% | #15 of 22 | Official leaderboard | |
| BioSecBench-Function | 31.8% | #19 of 27 | Official leaderboard | |
| TxBench-Oligo-Discovery | 31.9% | #22 of 30 | Official leaderboard | |
| SAFE-Teen Urgent safety | 51% | #13 of 17 | Official leaderboard | |
| xdev-bench-hard | 57.1% | #4 of 5 | Official leaderboard | |
| MetagenomicsBench | 27.3% | #21 of 26 | Official leaderboard | |
| SAFE-Teen Relationship boundaries | 64% | #14 of 17 | Official leaderboard | |
| BioSecBench-Surveillance | 36.6% | #23 of 27 | Official leaderboard | |
| scBench-Long | 10.2% | #24 of 27 | Official leaderboard | |
| TxBench-Antibody-Discovery | 26.3% | #27 of 30 | Official leaderboard | |
| TxBench-Preclinical-Pharmacology | 35% | #27 of 30 | Official leaderboard | |
| EpiBench | 12.9% | #25 of 27 | Official leaderboard | |
| SpatialBench-Long | 18.4% | #25 of 27 | Official leaderboard | |
| VariantBench | 26.3% | #25 of 27 | Official leaderboard | |
| SAFE-Teen Health advice | 70% | #16 of 17 | Official leaderboard | |
| CWE-bench 2 | 42% | #14 of 14 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LMCA | 55.5% | #17 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 63.4 | #28 of 151 | Official leaderboard | |
| SuperCLUE | 50.7 | #5 of 16 | Official leaderboard | |
| ACCoRD | 70.9 | #49 of 152 | Official leaderboard | |
| Hermes Index | 39.2 | #7 of 17 | Official leaderboard | |
| Xbench | 39.5 | #7 of 17 | Official leaderboard | |
| Political Science LLM Benchmark | 66.7% | #18 of 42 | Official leaderboard | |
| CAIS Risk Index | 48.8 | #29 of 67 | Official leaderboard | |
| CAIS Text Capabilities Index | 35.6 | #31 of 68 | Official leaderboard | |
| CompassRank | 74.4 | #11 of 14 | Official leaderboard |
The Self-Reported Index fits one score to the 8 capability results Z.ai 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 158.7 to 164.3. 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 |
|---|---|---|---|
| CyberGym | 84.5% CyberGym | 97.2% Leaderboard, dynamic, Sangfor AI | −13 pts |
| Terminal-Bench 3.0 | 28.3% Terminal Bench 3.0 | 32.4% Leaderboard, max, Claude Code | −4.1 pts |
| DeepSWE | 66.9% DeepSWE (v1.1) | 69% Leaderboard, max, mini-swe-agent | −2.1 pts |
| FrontierSWE v1 | 78.1% FrontierSWE | 78.1% Leaderboard, mean, Claude Code | Same |
Z.ai'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.