GLM-5.2 ranks first on 3 of the 206 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 | |
|---|---|---|---|---|
| StudyArena Text | 1,312 | #5 of 85 | Official leaderboard | |
| EQ-Bench 3 | 1,563 | #6 of 78 | Official leaderboard | |
| LMArena (Elo) | 1,475 | #31 of 335 | Official leaderboard | |
| Humanity's Last Exam | 54.7% | #27 of 154 | Lab model card | |
| Longform Creative Writing (EQ-Bench) | 77.9 | #23 of 119 | Official leaderboard | |
| Creative Writing v3 | 76.6 | #26 of 120 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 91111.0 | #20 of 77 | Official leaderboard | |
| SEA-HELM | 74.0 | #19 of 71 | Official leaderboard | |
| Judgemark v4 | 73.2 | #11 of 40 | Official leaderboard | |
| Phare | 69.9% | #21 of 62 | Official leaderboard | |
| Antidote | 1,040 | #15 of 36 | Official leaderboard | |
| Lech Mazur Creative Writing | 0.41 | #21 of 50 | Official leaderboard | |
| OpenAI MRCR v2 | 46.5% | #20 of 47 | Official leaderboard | |
| DesignArena ASCII Art | 1,186 | #33 of 77 | Official leaderboard | |
| Hemingway-bench | 1,036 | #18 of 42 | Official leaderboard | |
| Extended NYT Connections | 74.3% | #54 of 103 | Official leaderboard | |
| EQ-Bench 4 | 1,208 | #18 of 32 | Official leaderboard | |
| BullshitBench V2 | 31% | #77 of 129 | Official leaderboard | |
| HLE Extended | 25.8% | #19 of 31 | Official leaderboard | |
| SimpleQA Verified | 34.2% | #63 of 94 | Epoch AI run | |
| BullshitBench | 18.2% | #74 of 107 | Official leaderboard | |
| AA-LCR Extended | 42.7% | #39 of 49 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| IMO-AnswerBench | 91% | #1 of 37 | Lab model card | |
| GPQA Diamond | 91.9% | #31 of 229 | Epoch AI run | |
| DTBench | 93.6% | #32 of 180 | Official leaderboard | |
| HMMT Nov 2025 | 94.4% | #6 of 28 | Lab model card | |
| CritPt | 20.9% | #3 of 13 | Lab model card | |
| MathArena (Apex, IMO, USAMO) | 19.8% | #15 of 47 | Official leaderboard | |
| NatureBench | 13.3% | #6 of 18 | Official leaderboard | |
| ResearchClawBench | 20.7 | #8 of 23 | Official leaderboard | |
| Verifying the Verifiers (Project APE CRED benchmark) | 89.5% | #8 of 23 | Official leaderboard | |
| ProofBench | 35% | #22 of 59 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 86.4% | #84 of 212 | Epoch AI run | |
| HMMT 2026 | 92.4% | #13 of 31 | Official leaderboard | |
| SimpleBench | 58.8% | #47 of 106 | Official leaderboard | |
| BixBench3 | 46.3 | #9 of 20 | Official leaderboard | |
| Surface Evolver Bench | 55.6% | #12 of 25 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 59.2% | #42 of 86 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 29.3% | #34 of 68 | Official leaderboard | |
| MathArena Apex Shortlist | 68.1% | #19 of 38 | Official leaderboard | |
| ARC-AGI | 77% | #50 of 95 | Official leaderboard | |
| ARC-AGI-2 | 22.8% | #52 of 98 | Official leaderboard | |
| ASI-Bench | 35.6 | #7 of 11 | Official leaderboard | |
| Prophet Arena | 85.3 | #20 of 24 | Official leaderboard | |
| AIME 2026 | 90% | #28 of 31 | Official leaderboard | |
| xbench-ScienceQA | 64.6% | #29 of 29 | Official leaderboard | |
| FrontierChallenge | 3.10% | #11 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PostTrainBench | 34.3% | #1 of 23 | Official leaderboard | |
| SWE-bench Verified | 78.7% | #5 of 32 | Epoch AI run | |
| Data Agent Benchmark | 96.7% | #3 of 18 | Official leaderboard | |
| DesignArena Data Visualization | 1,283 | #20 of 118 | Official leaderboard | |
| WeirdML | 70.1% | #20 of 112 | Official leaderboard | |
| WebDev Arena | 1,593 | #21 of 105 | arena.ai, via Epoch AI | |
| DesignArena Website | 1,278 | #31 of 151 | Official leaderboard | |
| DesignArena UI Component | 1,288 | #29 of 139 | Official leaderboard | |
| SWE Atlas | 48.1% | #5 of 22 | Official leaderboard | |
| DesignArena 3D Design | 1,269 | #31 of 136 | Official leaderboard | |
| DesignArena Code | 1,278 | #33 of 143 | Official leaderboard | |
| Code Arena | 1,608 | #32 of 128 | Official leaderboard | |
| Monthly-SWEBench | 65% | #3 of 12 | Official leaderboard | |
| DesignArena Game Dev | 1,257 | #40 of 143 | Official leaderboard | |
| FrontierSWE v1 | 67.5% | #5 of 17 | Official leaderboard | |
| SWE-Bench Pro | 62.1% | #24 of 81 | Lab model card | |
| EvoCode-Bench v2 | 16.2% | #4 of 13 | Official leaderboard | |
| SWE-rebench | 62.9% | #4 of 13 | Official leaderboard | |
| Claw-SWE-Bench | 74.3% | #5 of 16 | Official leaderboard | |
| DesignArena Full Stack | 1,219 | #16 of 49 | Official leaderboard | |
| FrogsGame post-training task | 2.40% | #6 of 17 | Official leaderboard | |
| DesignArena SVG | 1,197 | #38 of 103 | Official leaderboard | |
| DesignArena Agentic Game Dev | 1,158 | #10 of 27 | Official leaderboard | |
| ALE-Bench | 1,685 | #48 of 126 | Official leaderboard | |
| SWE-Marathon | 13% | #5 of 12 | Official leaderboard | |
| DesignArena Agentic Overall | 1,187 | #32 of 74 | Official leaderboard | |
| NL2Repo-Bench | 48.9% | #13 of 30 | Lab model card | |
| DesignArena Frontend Web App | 1,204 | #23 of 53 | Official leaderboard | |
| Code Migration | 37.9% | #34 of 78 | Official leaderboard | |
| AndroidBench | 72.2% | #14 of 32 | Official leaderboard | |
| ProgramBench almost solved | 8.50% | #11 of 25 | Official leaderboard | |
| ProgramBench tests passed | 64.6% | #11 of 25 | Official leaderboard | |
| SWE Atlas Codebase QnA Extended | 16% | #17 of 38 | Official leaderboard | |
| SWE Atlas Test Writing | 41.5% | #10 of 22 | Official leaderboard | |
| Vibe Code Bench | 64% | #58 of 120 | Official leaderboard | |
| PostTrainBench 1.1 | 31.7% | #6 of 11 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,169 | #26 of 46 | Official leaderboard | |
| SWE Atlas Refactoring | 42.4% | #10 of 17 | Official leaderboard | |
| DesignArena React Native | 1,156 | #33 of 56 | Official leaderboard | |
| MLS-Bench | 40.4% | #9 of 15 | Official leaderboard | |
| SWE-bench Science | 31.9% | #6 of 10 | Official leaderboard | |
| SWE-bench Verified Extended | 47.3% | #27 of 44 | Official leaderboard | |
| APEX-SWE | 37.4% | #40 of 65 | Official leaderboard | |
| MirrorCode | 36.7% | #25 of 39 | Epoch AI run | |
| GameCraft-Bench | 39.1% | #10 of 15 | Official leaderboard | |
| ReactBench | 27.2% | #20 of 28 | Official leaderboard | |
| PostTrainBench 1.2 | 30.4% | #15 of 21 | Official leaderboard | |
| Tencent WorkBuddy Bench | 72.9% | #11 of 14 | Official leaderboard | |
| GameDevBench | 38.4% | #14 of 17 | Official leaderboard | |
| Senior SWE-Bench | 17.9% | #19 of 23 | Official leaderboard | |
| DeepSWE | 43.8% | #24 of 28 | Official leaderboard | |
| Next.js Agent Evals | 82% | #12 of 14 | Official leaderboard | |
| ProgramBench | 0% | #22 of 25 | Official leaderboard | |
| SWE-Together | 55% | #20 of 22 | Official leaderboard | |
| Doc2Harness | 27.6% | #18 of 19 | Official leaderboard | |
| GBAEval | 0.02% | #21 of 21 | Official leaderboard | |
| SWE Refactor Bench | 6.50% | #8 of 8 | Official leaderboard | |
| LoopArena | 24.2% | #6 of 6 | Official leaderboard | |
| Z.ai Code Bench | 23.4% | #5 of 5 | Official leaderboard | |
| ParallelKernelBench | 4.60% | #4 of 4 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ClawProBench | 80% | #1 of 49 | Official leaderboard | |
| GAIA | 89% | #2 of 64 | Official leaderboard | |
| PACT | 1,566 | #4 of 26 | Official leaderboard | |
| PinchBench 2.0 | 87% | #10 of 57 | Official leaderboard | |
| Pencil Puzzle Bench | 26.7% | #12 of 53 | Official leaderboard | |
| LiveClawBench | 72.9% | #4 of 17 | Official leaderboard | |
| BabaIsBench | 96% | #10 of 32 | Official leaderboard | |
| CivBench | 1,535 | #6 of 19 | Official leaderboard | |
| LisanBench | 1834.7 | #45 of 133 | Official leaderboard | |
| WildClawBench | 54.2% | #11 of 31 | Official leaderboard | |
| Terminal-Bench 2.1 | 82.7% | #23 of 56 | Lab model card | |
| τ³-bench Banking | 37.1% | #11 of 25 | Official leaderboard | |
| Harbor-Index | 9.80% | #4 of 9 | Official leaderboard | |
| Toolathlon-Verified | 59.9% | #12 of 25 | Official leaderboard | |
| MCP Atlas | 77.8% | #18 of 37 | Official leaderboard | |
| RuneBench | 4.76 | #37 of 76 | Official leaderboard | |
| clembench | 63.0 | #3 of 6 | Official leaderboard | |
| Long-Horizon-Terminal-Bench 2.0 | 50.3% | #8 of 15 | Official leaderboard | |
| BrowseComp Extended | 23% | #23 of 42 | Official leaderboard | |
| Toolathlon | 48.2% | #16 of 29 | Lab model card | |
| Agent Arena | 0.23% | #31 of 56 | Official leaderboard | |
| DuetBench | 1,009 | #5 of 9 | Official leaderboard | |
| CEO-Bench | $54,327 | #11 of 19 | Official leaderboard | |
| Long-Horizon-Terminal-Bench (LHTB) | 31.6% | #15 of 25 | Official leaderboard | |
| Mystery Game Puzzles | 19% | #53 of 84 | Epoch AI run | |
| Chess Puzzles | 22% | #63 of 93 | Epoch AI run | |
| Terminal-Bench 2.1 Extended | 27.9% | #31 of 44 | Official leaderboard | |
| AutomationBench | 7.76% | #30 of 42 | Official leaderboard | |
| DiG-bench | 15.2% | #5 of 7 | Official leaderboard | |
| CAR-bench | 43.3% | #10 of 12 | Official leaderboard | |
| EBR-bench | 9.52% | #27 of 31 | Epoch AI run | |
| Terminal-Bench 3.0 | 4.59% | #11 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CorpFin v2 | 66.1% | #22 of 119 | Official leaderboard | |
| FORTE | 76.9% | #4 of 21 | Official leaderboard | |
| DesignArena PDF | 1,243 | #8 of 37 | Official leaderboard | |
| Vending-Bench 2 | $8,314 | #16 of 67 | Official leaderboard | |
| LegalBench | 84.1% | #40 of 138 | Official leaderboard | |
| LEXam-hard | 37.5% | #5 of 15 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 7.08% | #29 of 80 | Official leaderboard | |
| E-Commerce Bench | 3.01 | #7 of 19 | Official leaderboard | |
| TaxEval v2 | 73.3% | #48 of 129 | Official leaderboard | |
| MerchantBench | 42.3 | #3 of 8 | Official leaderboard | |
| Excel Modeling Benchmark | 61.5% | #33 of 76 | Official leaderboard | |
| MedScribe | 83.5% | #44 of 97 | Official leaderboard | |
| Vals Index v2 | 53.1% | #31 of 65 | Official leaderboard | |
| Physical-Engineering's Last Exam | 0.25 | #9 of 18 | Official leaderboard | |
| CHI-Bench | 18.7% | #12 of 23 | Official leaderboard | |
| DesignArena Python-PPTX Slides | 1,159 | #19 of 36 | Official leaderboard | |
| JobBench | 43.4% | #9 of 16 | Official leaderboard | |
| ProfBench | 45.8% | #41 of 72 | Official leaderboard | |
| Vals Finance Agent v2 | 49.7% | #46 of 80 | Official leaderboard | |
| YC-Bench | $1,013,158 | #34 of 59 | Official leaderboard | |
| Legal Research Bench | 31.3% | #54 of 93 | Official leaderboard | |
| MedCode | 40.8% | #57 of 96 | Official leaderboard | |
| DesignArena HTML Slides | 1,071 | #17 of 28 | Official leaderboard | |
| Agents' Last Exam | 20.4% | #20 of 31 | Official leaderboard | |
| prinzbench | 30.0 | #17 of 26 | Official leaderboard | |
| APEX-1 | 61.5% | #48 of 71 | Official leaderboard | |
| CommerceAgentBench | 46.7% | #9 of 13 | Official leaderboard | |
| BankerToolBench v1.1 | 12.4% | #7 of 10 | Official leaderboard | |
| APEX-Agents 1.1 | 61.1% | #46 of 65 | Official leaderboard | |
| APEX-Agents | 45.2% | #44 of 61 | Collected by Epoch AI, source not named | |
| GDPevo | 60.1% | #3 of 4 | Official leaderboard | |
| DesignArena Spreadsheet | 1,064 | #28 of 35 | Official leaderboard | |
| TRLawBench | 8.02 | #28 of 34 | Official leaderboard | |
| FrontierFinance | 42.8% | #14 of 17 | Official leaderboard | |
| PolicyBench | 76.5% | #42 of 50 | Official leaderboard | |
| GDPval Extended | 0.20% | #38 of 43 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,547 | #36 of 75 | Official leaderboard | |
| ArtifactArena | 1,051 | #13 of 21 | Official leaderboard | |
| VoxelBench | 1,585 | #36 of 56 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ABC Bench Fragment Design | 99.8% | #3 of 27 | Official leaderboard | |
| Human Pathogen Capabilities Test | 65.3% | #13 of 86 | Official leaderboard | |
| ABC Bench Screening Evasion | 70.4% | #3 of 17 | Official leaderboard | |
| ABC Bench Liquid-Handling Robot | 99.3% | #7 of 34 | Official leaderboard | |
| CyberGym | 90.8% | #7 of 34 | Official leaderboard | |
| Molecular Biology Capabilities Test | 55.5% | #19 of 90 | Official leaderboard | |
| CyberBench (Vals) | 77.4% | #7 of 26 | Official leaderboard | |
| Agentic BAIM-LLM Evaluation (ABLE) | 76.7% | #5 of 18 | Official leaderboard | |
| BioTIER Permit | 99.7% | #23 of 69 | Official leaderboard | |
| nocot-bench (NCRI) | 77.3 | #65 of 185 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 59.6% | #29 of 71 | Official leaderboard | |
| World-Class Bio | 47% | #36 of 74 | Official leaderboard | |
| BioTIER Refuse | 32.6% | #42 of 69 | Official leaderboard | |
| Dictatorship Eval | 46.6% | #13 of 20 | Official leaderboard | |
| DrugDiscoveryBench | 36.2% | #13 of 20 | Official leaderboard | |
| PatchEval | 67% | #5 of 7 | Official leaderboard | |
| PatchEval-Verified | 67% | #5 of 7 | Official leaderboard | |
| PCT (Polarized Contrastive Pairs) | 50% | #33 of 46 | Official leaderboard | |
| ExploitGym | 4.49% | #13 of 18 | Official leaderboard | |
| GeneBench Pro | 4.64% | #9 of 12 | Official leaderboard | |
| FORTRESS | 48.7% | #57 of 73 | Official leaderboard | |
| DeepsecBench | 10.4 | #25 of 29 | Official leaderboard | |
| dfbench | 40.7% | #16 of 18 | Official leaderboard | |
| SRE-Bench | 0% | #33 of 35 | Official leaderboard | |
| HackBench | 32.3 | #8 of 8 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LMCA | 45.8% | #39 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 58.6 | #44 of 151 | Official leaderboard | |
| CAIS Text Capabilities Index | 37.4 | #27 of 68 | Official leaderboard | |
| ACCoRD | 66.0 | #68 of 152 | Official leaderboard | |
| CAIS Risk Index | 51.9 | #37 of 67 | Official leaderboard |
The Self-Reported Index fits one score to the 14 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 151.3 to 157.6. 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 |
|---|---|---|---|
| AIME 2026 | 99.2% AIME 2026 | 90% Leaderboard, max | +9.2 pts |
| FrontierSWE v1 | 74.4% FrontierSWE (Dominance) | 67.5% Leaderboard, mean, Claude Code | +6.9 pts |
| DeepSWE | 46.2% DeepSWE | 43.8% Leaderboard, max, mini-swe-agent | +2.4 pts |
| MCP Atlas | 76.8% MCP-Atlas (Public Set) | 77.8% Leaderboard | −1.0 pts |
| GPQA Diamond | 91.2% GPQA-Diamond | 91.9% Epoch AI run, max | −0.7 pts |
| HMMT 2026 | 92.5% HMMT Feb. 2026 | 92.4% Leaderboard, max | Same |
| PostTrainBench | 34.3% PostTrainBench | 34.3% Leaderboard, max, Claude Code | Same |
| SWE-Marathon | 13% SWE-Marathon | 13% Leaderboard, 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.