- Mean of TAU-bench-Retail / TAU-bench-Airline 70
- BFCL v3 (Full) 77.8
- BrowseComp 26.4
- MMLU Pro 84.6
- AIME24 91
- MATH 500 98.2
- SciCode 41.7
- GPQA 79.1
- HLE 14.4
- LiveCodeBench (2407-2501)
- AA-Index (Estimated) 67.7
- SWE-bench Verified 64.2
- SWE-Terminal-Bench (Terminal Bench) 37.5
GLM-4.5
GLM-4.5 places in the top three on 8 benchmarks.
93
Benchmarks with a score
0
Ranked first, out of five or more models
8
In the top three
Z.ai Aug 2025
GLM-4.5
Open weights Flagship
136.2
#219 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- Humanity's Last Exam 8.32% #31 /47
- SciCode 41.7% #26 /49
- MathArena (Apex, IMO, USAMO) 1.04% #37 /47
Where it ranks
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.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM Confabulation Leaderboard | 11.3% | #5 of 58 | Official leaderboard | |
| MMLU | 90% | #16 of 131 | Lab model card | |
| PIQA | 85.3% | #5 of 37 | Lab model card | |
| MMLU-Pro | 84.6% | #36 of 167 | Lab model card | |
| BIG-Bench Hard | 86.2% | #22 of 73 | Lab model card | |
| C3 | 83.1% | #2 of 6 | Lab model card | |
| SysBench | 81% | #2 of 6 | Lab model card | |
| LMArena (Elo) | 1,411 | #119 of 335 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 16000.0 | #29 of 77 | Official leaderboard | |
| HellaSwag | 87.1% | #20 of 53 | Lab model card | |
| Vectara Hallucination Leaderboard (HHEM) | 3.70% | #58 of 148 | Official leaderboard | |
| LongBench v2 | 50.3% | #13 of 33 | Official leaderboard | |
| Meeseeks Utility Rate | 86.1% | #4 of 10 | Official leaderboard | |
| EQ-Bench 3 | 1,322 | #34 of 78 | Official leaderboard | |
| SimpleQA | 26.4% | #28 of 60 | Lab model card | |
| TriviaQA | 80% | #19 of 40 | Lab model card | |
| IFEval | 86.1% | #72 of 135 | Lab model card | |
| ConStory-Bench | 0.59 | #8 of 14 | Official leaderboard | |
| C-Eval | 86.9% | #28 of 46 | Lab model card | |
| Chinese SimpleQA | 70.1% | #8 of 13 | Lab model card | |
| Creative Writing v3 | 63.0 | #75 of 120 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 55.5 | #77 of 119 | Official leaderboard | |
| Slop Score | 59.9 | #11 of 17 | Official leaderboard | |
| Humanity's Last Exam | 8.32% | #31 of 47 | Official leaderboard | |
| Judgemark v2.1 | 63.4 | #29 of 42 | Official leaderboard | |
| MultiChallenge | 52.8% | #30 of 42 | Lab model card | |
| BullshitBench | 14.6% | #82 of 107 | Official leaderboard | |
| MultiNRC | 17.4% | #33 of 38 | Official leaderboard | |
| CLUEWSC | 83.5% | #7 of 8 | Lab model card | |
| BullshitBench V2 | 8.00% | #119 of 129 | Official leaderboard | |
| Challenging translation human evaluation (Z.ai) | 1.71 | #1 of 1 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH-500 | 98.2% | #4 of 70 | Lab model card | |
| AIME 2024 | 91% | #15 of 88 | Lab model card | |
| AIME 2025 | 93.3% | #11 of 53 | Official leaderboard | |
| AMO-Bench | 36.8% | #7 of 21 | Official leaderboard | |
| GPQA Diamond | 79.1% | #123 of 302 | Lab model card | |
| BRUMO 2025 | 92.5% | #18 of 42 | Official leaderboard | |
| PHYBench | 26% | #13 of 28 | Official leaderboard | |
| HMMT 2025 | 77.5% | #25 of 53 | Official leaderboard | |
| Novel logical reasoning problems expert evaluation (Z.ai) | 62.0 | #1 of 2 | Lab model card | |
| SciArena | 1,007 | #19 of 34 | Official leaderboard | |
| MATH | 61% | #59 of 99 | Lab model card | |
| SMT 2025 | 82.1% | #28 of 40 | Official leaderboard | |
| CMIMC 2025 | 72.5% | #24 of 33 | Official leaderboard | |
| GSM8K | 79.4% | #67 of 92 | Lab model card | |
| MathArena (Apex, IMO, USAMO) | 1.04% | #37 of 47 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v5 | 72.9% | #5 of 47 | Lab model card | |
| AutoCodeBench | 46.6% | #12 of 34 | Official leaderboard | |
| ArtifactsBench | 51.3% | #8 of 20 | Official leaderboard | |
| SWE-bench Verified | 64.2% | #25 of 60 | Official leaderboard | |
| AlgoTune | 1.52× | #9 of 18 | Official leaderboard | |
| CC-Bench | 36.5% | #2 of 4 | Official leaderboard | |
| SciCode | 41.7% | #26 of 49 | Lab model card | |
| DesignArena 3D Design | 1,153 | #73 of 136 | Official leaderboard | |
| CATArena Gomoku | 905 | #3 of 5 | Official leaderboard | |
| CATArena OIBench-5 | 0.18 | #3 of 5 | Official leaderboard | |
| CATArena SWE-Perf-10 | 0.68× | #3 of 5 | Official leaderboard | |
| DesignArena Game Dev | 1,135 | #89 of 143 | Official leaderboard | |
| DesignArena Website | 1,163 | #94 of 151 | Official leaderboard | |
| DesignArena Code | 1,155 | #93 of 143 | Official leaderboard | |
| WeirdML | 40.6% | #76 of 112 | Official leaderboard | |
| DesignArena UI Component | 1,135 | #95 of 139 | Official leaderboard | |
| DAXBench v2 | 71% | #33 of 47 | Official leaderboard | |
| DesignArena SVG | 1,086 | #75 of 103 | Official leaderboard | |
| DesignArena Data Visualization | 1,157 | #88 of 118 | Official leaderboard | |
| CATArena Texas Holdem | 972 | #4 of 5 | Official leaderboard | |
| EvalPlus | 78.1% | #5 of 6 | Lab model card | |
| IOI (Vals) | 2.92% | #47 of 55 | Official leaderboard | |
| ALE-Bench | 552 | #125 of 126 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Berkeley Function Calling Leaderboard (BFCL) | 77.8% | #4 of 41 | Lab model card | |
| C3-Bench | 12.1% | #3 of 30 | Official leaderboard | |
| tau-bench | 70.1% | #3 of 23 | Lab model card | |
| MCP-Bench | 0.67 | #5 of 15 | Official leaderboard | |
| Elimination Game | 4 | #27 of 56 | Official leaderboard | |
| LisanBench | 400.3 | #71 of 133 | Official leaderboard | |
| CompileBench | 73.3% | #11 of 20 | Official leaderboard | |
| Step Race | 2 | #33 of 59 | Official leaderboard | |
| MCP-Universe | 24.7% | #17 of 30 | Official leaderboard | |
| VitaBench 2.0 | 36.4% | #11 of 17 | Official leaderboard | |
| Terminal-Bench | 37.5% | #15 of 22 | Lab model card | |
| MCPMark | 15.6% | #22 of 28 | Official leaderboard | |
| SnakeBench | 17 | #161 of 188 | Official leaderboard | |
| lmgame-Bench | 11.7 | #21 of 23 | Official leaderboard | |
| BrowseComp | 26.4% | #80 of 81 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| StockBench | 3.00 | #3 of 13 | Official leaderboard | |
| TaxEval v2 | 72.4% | #61 of 129 | Official leaderboard | |
| CorpFin v2 | 61% | #62 of 119 | Official leaderboard | |
| LegalBench | 75.6% | #105 of 138 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 71.3 | #92 of 185 | Official leaderboard | |
| SafetyBench | 89.9% | #1 of 2 | Lab model card | |
| Lakera B3 (Backbone Breaker Benchmark) | 48.6% | #28 of 35 | Official leaderboard | |
| FORTRESS | 59.6% | #68 of 73 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Artificial Analysis Intelligence Index v2 | 67.7 | #1 of 2 | Lab model card |
Its lab's numbers put it at 136.2, number 219 of 404
The Self-Reported Index fits one score to the 25 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 133.7 to 138.6. How we compute it.
- MATH-500 98.2
- AIME 2024 91
- MMLU 90
- HellaSwag 87.1
- C-Eval 86.9
- BIG-Bench Hard 86.2
- IFEval 86.1
- PIQA 85.3
- MMLU-Pro 84.6
- CLUEWSC 83.5
- TriviaQA 80
- GSM8K 79.4
- GPQA Diamond 79.1
- Berkeley Function Calling Leaderboard (BFCL) 77.8
- LiveCodeBench v5 72.9
- tau-bench 70.1
- Chinese SimpleQA 70.1
- SWE-bench Verified 64.2
- MATH 61
- MultiChallenge 52.8
- SciCode 41.7
- Terminal-Bench 37.5
- BrowseComp 26.4
- SimpleQA 26.4
- Humanity's Last Exam 14.4
Its lab's numbers next to 4 leaderboards
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.
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 22 of them aren't in the launch post.
Capabilities
- tau-bench 70.1
- Berkeley Function Calling Leaderboard (BFCL) 77.8
- BrowseComp 26.4
- MMLU-Pro 84.6
- AIME 2024 91
- MATH-500 98.2
- SciCode 41.7
- GPQA Diamond 79.1
- Humanity's Last Exam 14.4
- LiveCodeBench v5 72.9
- Artificial Analysis Intelligence Index v2 67.7
- SWE-bench Verified 64.2
- Terminal-Bench 37.5
- MMLU 90
- IFEval 86.1
- SysBench 81
- MultiChallenge 52.8
- Novel logical reasoning problems expert evaluation (Z.ai) 62
- Real-scenario prompts human evaluation (Z.ai)
- CC-Bench
- Challenging translation human evaluation (Z.ai) 1.71
- BIG-Bench Hard 86.2
- HellaSwag 87.1
- PIQA 85.3
- TriviaQA 80
- EvalPlus 78.1
- GSM8K 79.4
- MATH 61
- CLUEWSC 83.5
- C-Eval 86.9
- C3 83.1
Hallucination
- SimpleQA 26.4
- Chinese SimpleQA 70.1
Safety
- SafetyBench 89.9