GLM-5 ranks first on EvalPlus.
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 | |
|---|---|---|---|---|
| Judgemark v2.1 | 85.6 | #4 of 42 | Official leaderboard | |
| MMLU-Pro | 86% | #27 of 212 | Official leaderboard | |
| LongBench v2 | 64.5% | #5 of 35 | Lab model card | |
| LMArena (Elo) | 1,457 | #51 of 335 | Official leaderboard | |
| MMLU | 88.3% | #26 of 131 | Lab model card | |
| BIG-Bench Hard | 87.4% | #15 of 73 | Lab model card | |
| HellaSwag | 88.1% | #13 of 53 | Lab model card | |
| Humanity's Last Exam | 50.4% | #39 of 154 | Lab model card | |
| MultiChallenge | 64.8% | #11 of 42 | Lab model card | |
| PIQA | 84.6% | #10 of 37 | Lab model card | |
| EQ-Bench 3 | 1,495 | #24 of 78 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 70.9 | #39 of 119 | Official leaderboard | |
| SimpleQA | 36.9% | #20 of 60 | Lab model card | |
| KINA | 35.9% | #14 of 37 | Official leaderboard | |
| Creative Writing v3 | 70.6 | #50 of 120 | Official leaderboard | |
| IFBench | 72% | #34 of 80 | Lab model card | |
| TriviaQA | 80.9% | #17 of 40 | Lab model card | |
| CL-bench | 18.7% | #11 of 24 | Official leaderboard | |
| Chinese SimpleQA | 75.2% | #6 of 13 | Lab model card | |
| SkateBench v2 | 66.7% | #10 of 21 | Official leaderboard | |
| C-Eval | 88.8% | #23 of 46 | Lab model card | |
| C3 | 80.3% | #3 of 6 | Lab model card | |
| Extended NYT Connections | 74.8% | #52 of 103 | Official leaderboard | |
| AraGen | 46% | #23 of 42 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 10.1% | #60 of 104 | Official leaderboard | |
| BullshitBench V2 | 28% | #81 of 129 | Official leaderboard | |
| DesignArena ASCII Art | 1,142 | #50 of 77 | Official leaderboard | |
| BullshitBench | 20% | #71 of 107 | Official leaderboard | |
| CLUEWSC | 84.2% | #6 of 8 | Lab model card | |
| HLE Extended | 2.00% | #31 of 31 | Official leaderboard | |
| LOCA-Bench | 4.00% | #9 of 9 | Official leaderboard | |
| IF-Badcase | 83.2% | #1 of 1 | Lab model card | |
| MENT-SNS | 1013.0 | #1 of 1 | Lab model card | |
| Slide generation human evaluation (Z.ai) | 67.5% | #1 of 1 | Lab model card | |
| Translation Human Scoring (Z.ai) | 3.12 | #1 of 1 | Lab model card | |
| ZMultiDialBench | 8.57 | #1 of 1 | Lab model card | |
| ZMultiTransBench | 1050.0 | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 96.7% | #4 of 53 | Official leaderboard | |
| BRUMO 2025 | 99.2% | #4 of 42 | Official leaderboard | |
| HMMT 2025 | 97.9% | #6 of 54 | Lab model card | |
| General365 | 59.9% | #3 of 23 | Official leaderboard | |
| HMMT Nov 2025 | 94.2% | #3 of 22 | Official leaderboard | |
| CMIMC 2025 | 93.1% | #5 of 33 | Official leaderboard | |
| SMT 2025 | 91% | #8 of 40 | Official leaderboard | |
| AIME 2026 | 96.7% | #9 of 31 | Official leaderboard | |
| GPQA Diamond | 87.8% | #67 of 229 | Epoch AI run | |
| MathArena (Apex, IMO, USAMO) | 10.9% | #21 of 47 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 80% | #97 of 212 | Epoch AI run | |
| MathArena Apex Shortlist | 68.6% | #18 of 38 | Official leaderboard | |
| FrontierMath | 16.4% | #27 of 55 | Official leaderboard | |
| ForecastBench | 0.13 | #42 of 82 | Official leaderboard | |
| IMO-AnswerBench | 82.5% | #19 of 37 | Lab model card | |
| SimpleBench | 53.2% | #55 of 106 | Official leaderboard | |
| ChemBench | 59.3% | #13 of 24 | Official leaderboard | |
| KellyBench | -51.60% | #3 of 5 | Official leaderboard | |
| MATH | 56.4% | #62 of 99 | Lab model card | |
| FrontierMath Tier 4 | 2.10% | #38 of 55 | Epoch AI run | |
| ARC-AGI-2 | 4.86% | #70 of 98 | Official leaderboard | |
| ARC-AGI | 44.7% | #70 of 95 | Official leaderboard | |
| HMMT 2026 | 86.4% | #23 of 31 | Official leaderboard | |
| Echo General AI Prediction Leaderboard | 984 | #13 of 17 | Official leaderboard | |
| GSM8K | 68.8% | #73 of 92 | Lab model card | |
| USAMO 2026 | 35.1% | #9 of 9 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| EvalPlus | 87% | #1 of 6 | Lab model card | |
| IOI (Vals) | 22% | #20 of 55 | Official leaderboard | |
| DesignArena Game Dev | 1,224 | #54 of 143 | Official leaderboard | |
| Spider 2.0 | 50.3% | #9 of 23 | Official leaderboard | |
| DesignArena Code | 1,237 | #56 of 143 | Official leaderboard | |
| DesignArena 3D Design | 1,212 | #54 of 136 | Official leaderboard | |
| DesignArena Website | 1,241 | #62 of 151 | Official leaderboard | |
| DesignArena Data Visualization | 1,226 | #52 of 118 | Official leaderboard | |
| WeirdML | 48.2% | #51 of 112 | Official leaderboard | |
| DesignArena UI Component | 1,222 | #66 of 139 | Official leaderboard | |
| SWE-bench Multilingual | 73.3% | #35 of 73 | Lab model card | |
| AutoLab | 0.43 | #5 of 10 | Official leaderboard | |
| OJBench | 54.9% | #4 of 8 | Official leaderboard | |
| CC-Bench-V2 Backend | 25.8% | #1 of 2 | Lab model card | |
| CC-Bench-V2 Frontend | 74.8% | #1 of 2 | Lab model card | |
| OpenHands Index | 49.4% | #17 of 33 | Official leaderboard | |
| DesignArena SVG | 1,159 | #55 of 103 | Official leaderboard | |
| WebDev Arena | 1,436 | #59 of 105 | arena.ai, via Epoch AI | |
| Code Arena | 1,435 | #76 of 128 | Official leaderboard | |
| DesignArena Agentic Overall | 1,108 | #49 of 74 | Official leaderboard | |
| ALE-Bench | 1,194 | #84 of 126 | Official leaderboard | |
| DesignArena Full Stack | 1,101 | #33 of 49 | Official leaderboard | |
| SWE-bench Verified | 72.1% | #22 of 32 | Epoch AI run | |
| Vibe Code Bench | 23.4% | #83 of 120 | Official leaderboard | |
| PostTrainBench | 13.9% | #16 of 23 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,135 | #33 of 46 | Official leaderboard | |
| DesignArena React Native | 1,130 | #41 of 56 | Official leaderboard | |
| DAXBench v2 | 66% | #37 of 47 | Official leaderboard | |
| SWE Atlas | 20.5% | #18 of 22 | Official leaderboard | |
| SWE Atlas Refactoring | 24.2% | #14 of 17 | Official leaderboard | |
| SWE Atlas Test Writing | 28.7% | #19 of 22 | Official leaderboard | |
| CC-Bench-V2 Large Repo Exploration | 65.6% | #2 of 2 | Lab model card | |
| CC-Bench-V2 Multi-Step Chained Tasks | 52.3% | #1 of 1 | Lab model card | |
| SWE-rebench | 42.1% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 31 | #36 of 188 | Official leaderboard | |
| PinchBench | 76.2% | #10 of 48 | Official leaderboard | |
| tau2-bench | 89.7% | #17 of 79 | Lab model card | |
| Claw-Eval-Live | 61.9% | #3 of 12 | Official leaderboard | |
| BrowseComp-ZH | 72.7% | #5 of 19 | Lab model card | |
| Terminal-Bench 2.0 | 52.4% | #16 of 43 | Official leaderboard | |
| Toolathlon | 39.2% | #20 of 50 | Official leaderboard | |
| LisanBench | 986.8 | #56 of 133 | Official leaderboard | |
| BrowseComp | 75.9% | #40 of 81 | Lab model card | |
| WildClawBench | 46.4% | #19 of 31 | Official leaderboard | |
| QwenClawBench | 54.9% | #8 of 12 | Official leaderboard | |
| ClawProBench | 57% | #33 of 49 | Official leaderboard | |
| MCP Atlas | 67.8% | #32 of 46 | Lab model card | |
| EnvSimBench | 41% | #5 of 7 | Official leaderboard | |
| MM-BrowseComp | 3.25% | #8 of 11 | Official leaderboard | |
| LLM Chess | -299 | #101 of 123 | Official leaderboard | |
| tau3-bench | 81% | #6 of 7 | Official leaderboard | |
| DeepPlanning | 14.6% | #13 of 15 | Official leaderboard | |
| Pencil Puzzle Bench | 0% | #46 of 53 | Official leaderboard | |
| RuneBench | 1.90 | #71 of 76 | Official leaderboard | |
| τ³-bench Banking | 9.79% | #24 of 25 | Official leaderboard | |
| ToolCall-Badcase | 95.8% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LegalBench | 84.1% | #42 of 138 | Official leaderboard | |
| Vals Index v1 | 55.1% | #13 of 34 | Official leaderboard | |
| StrataBench Land v1 Extract | 88% | #2 of 5 | Official leaderboard | |
| CorpFin v2 | 62.9% | #48 of 119 | Official leaderboard | |
| YC-Bench | $1,208,190 | #25 of 59 | Official leaderboard | |
| Vals Finance Agent | 53.2% | #21 of 49 | Official leaderboard | |
| Vending-Bench Arena | $5,582 | #16 of 32 | Official leaderboard | |
| SpreadsheetBench 2 | 17.1% | #4 of 8 | Official leaderboard | |
| Vending-Bench 2 | $4,432 | #39 of 67 | Official leaderboard | |
| AlphaEval | 49.9% | #3 of 5 | Official leaderboard | |
| TaxEval v2 | 70% | #88 of 129 | Official leaderboard | |
| OccuBench | 62.6% | #10 of 14 | Official leaderboard | |
| ProfBench | 43.3% | #52 of 72 | Official leaderboard | |
| DesignArena HTML Slides | 1,043 | #22 of 28 | Official leaderboard | |
| FORTE | 60.7% | #17 of 21 | Official leaderboard | |
| CaseLaw v2 | 52.5% | #45 of 53 | Official leaderboard | |
| GDPval-AA | 1,409 | #14 of 16 | Lab model card | |
| BankerToolBench v1.1 | 4.70% | #9 of 10 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 17.4% | #46 of 51 | Official leaderboard | |
| FoodTruck Bench | $-210 | #57 of 59 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,214 | #55 of 75 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BioTIER Permit | 99.7% | #21 of 69 | Official leaderboard | |
| nocot-bench (NCRI) | 76.2 | #72 of 185 | Official leaderboard | |
| LiveSecBench | 56.7 | #16 of 34 | Official leaderboard | |
| Human Pathogen Capabilities Test | 53.4% | #41 of 86 | Official leaderboard | |
| SusVibes | 14% | #5 of 10 | Official leaderboard | |
| KORA | 40.6% | #22 of 42 | Official leaderboard | |
| DistressBench | 71.2% | #23 of 42 | Official leaderboard | |
| Molecular Biology Capabilities Test | 48% | #50 of 90 | Official leaderboard | |
| BioTIER Refuse | 33.9% | #40 of 69 | Official leaderboard | |
| CyberGym | 43.2% | #20 of 34 | Official leaderboard | |
| BinaryAudit | 19.3% | #12 of 20 | Official leaderboard | |
| SEC-bench Pro | 3.80% | #4 of 6 | Official leaderboard | |
| SEC-bench Pro 260617 | 3.80% | #4 of 6 | Official leaderboard | |
| World-Class Bio | 38.4% | #50 of 74 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Artificial Analysis Intelligence Index v4.0 | 50.0 | #2 of 5 | Lab model card |
The Self-Reported Index fits one score to the 28 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 144.5 to 151.3. 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 | What differs |
|---|---|---|---|---|
| SWE-bench Verified | 77.8% SWE-bench Verified | 72.8% Leaderboard, high, mini-SWE-agent | +5.0 pts | |
| AIME 2026 | 92.7% AIME 2026 I | 96.7% Leaderboard | −4.0 pts | |
| Terminal-Bench 2.0 | 56.2% Terminal-Bench 2.0 (Terminus 2) | 52.4% Leaderboard, Terminus 2 | +3.8 pts | |
| SWE-bench Multilingual | 73.3% SWE-bench Multilingual | 69.7% Leaderboard, mini-SWE-agent | +3.6 pts | |
| HMMT Nov 2025 | 96.9% HMMT Nov. 2025 | 94.2% Leaderboard | +2.7 pts | |
| GPQA Diamond | 86% GPQA-Diamond | 87.8% Epoch AI run | −1.8 pts | |
| Code Arena | 1,449 LMArena Code Arena (Agentic WebDev) score, rank 6 (Figure 3 screenshot) | 1,435 Leaderboard | +14 | The lab reproduces an early Code Arena snapshot, while the board uses a later rating from additional votes. |
| HMMT 2025 | 97.9% HMMT Feb. 2025 (Table 7) | 97.5% Leaderboard | +0.4 pts | |
| LMArena (Elo) | 1,452 LMArena Text Arena score, rank 11 (Figure 3 screenshot); LMsys Chatbot Arena 1452 (Figure 11) | 1,457 Leaderboard | −5 | Z.ai quotes an earlier Arena rating, while the board reports a later rating from updated votes and scoring. |
| CC-Bench-V2 Backend | 25.8% CC-Bench-V2 Backend Engineering, Pass@1 (Table 8) | 25.8% Leaderboard | Same | |
| CC-Bench-V2 Multi-Step Chained Tasks | 52.3% CC-Bench-V2 Long-horizon Chained Tasks, Pass@1 (Table 8) | 52.3% Leaderboard | Same | |
| CC-Bench-V2 Frontend | 74.8% CC-Bench-V2 Frontend, End-to-End Correctness (Figure 4). Build Success Rate 98.0; per-stack ISR/CSR in Table 8 | 74.8% Leaderboard | Same | |
| CC-Bench-V2 Large Repo Exploration | 65.6% CC-Bench-V2 Long-horizon Repo Exploration, Pass@1 (Table 8) | 65.6% Leaderboard | Same | |
| CyberGym | 43.2% CyberGym | 43.2% Leaderboard, Claude Code | Same | |
| Toolathlon | 39.2% Tool-Decathlon (Table 7) | 39.2% Leaderboard | Same | |
| Vending-Bench 2 | $4,432 Vending Bench 2 | $4,432 Leaderboard | 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.
Every result in the longer documents the lab published with it. 34 of them aren't in the launch post.
