Apertus 70B
Apertus 70B ranks first on BLEnD.
Swiss AI Initiative Released September 2nd, 2025 Family Apertus The lab's model card Compare with others 37
Benchmarks with a score
1
Ranked first, out of five or more models
1
In the top three
Swiss AI Initiative Sep 2025 Apertus 70B
Open weights Flagship
110.3
#377 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- LEXam 34.7% #23 /30
- LLM-WikiRace 0% #20 /21
- CulturalRiddles 15.2% #10 /20
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 | |
|---|---|---|---|---|
| BLEnD | 66.3% | #1 of 5 | Lab model card | |
| Global PIQA | 74.3% | #15 of 44 | Official leaderboard | |
| Multi-IF | 74.7% | #11 of 28 | Lab model card | |
| TruthfulQA | 61.2% | #14 of 30 | Lab model card | |
| ARC-Challenge | 85% | #23 of 49 | Lab model card | |
| CulturalRiddles | 15.2% | #10 of 20 | Official leaderboard | |
| PIQA | 81.9% | #22 of 37 | Lab model card | |
| WinoGrande | 73.3% | #34 of 47 | Lab model card | |
| MMLU | 69.6% | #101 of 131 | Lab model card | |
| BIG-Bench Hard | 64.2% | #57 of 73 | Lab model card | |
| HellaSwag | 78.1% | #42 of 53 | Lab model card | |
| MGSM | 64.3% | #47 of 57 | Lab model card | |
| INCLUDE | 58.2% | #26 of 31 | Lab model card | |
| IFEval | 75.2% | #116 of 135 | Lab model card | |
| Last Translation Benchmark | 1.21 | #31 of 36 | Official leaderboard | |
| AGIEval | 40.5% | #35 of 40 | Lab model card | |
| Global MMLU | 62.7% | #13 of 14 | Lab model card | |
| SEA-HELM | 32.6 | #67 of 71 | Official leaderboard | |
| DROP | 50.8% | #40 of 41 | Lab model card | |
| MLogiQA | 40.5% | #8 of 8 | Lab model card | |
| XCOPA | 69.8% | #2 of 2 | Lab model card | |
| ACPBench | 62.9% | #1 of 1 | Lab model card | |
| CulturalBench | 74.2% | #1 of 1 | Lab model card | |
| SwitzerlandQA | 67.2% | #1 of 1 | Lab model card | |
| XNLI | 45.3% | #1 of 1 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GSM8K | 77.6% | #68 of 92 | Lab model card | |
| MATH | 30.8% | #86 of 99 | Lab model card | |
| GPQA Main | 30.6% | #7 of 7 | Lab model card | |
| GSM8K-Platinum | 74.6% | #1 of 1 | Lab model card | |
| MathQA | 33.9% | #1 of 1 | Lab model card |
Coding
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM-WikiRace | 0% | #20 of 21 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LEXam | 34.7% | #23 of 30 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| RealToxicityPrompts | 0.20% | #1 of 2 | Lab model card | |
| BBQ (Bias Benchmark for QA) | 67.4% | #23 of 25 | Lab model card | |
| HarmBench | 31.9% | #5 of 5 | Lab model card |
Its lab's numbers put it at 110.3, number 377 of 404
The Self-Reported Index fits one score to the 19 capability results Swiss AI Initiative 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 107.7 to 112.9. How we compute it.
- ARC-Challenge 85
- PIQA 81.9
- HellaSwag 78.1
- GSM8K 77.6
- IFEval 75.2
- Multi-IF 74.7
- WinoGrande 73.3
- HumanEval 73
- MMLU 69.6
- MGSM 64.3
- BIG-Bench Hard 64.2
- Global MMLU 62.7
- TruthfulQA 61.2
- INCLUDE 58.2
- DROP 50.8
- MBPP 47
- AGIEval 40.5
- MLogiQA 40.5
- MATH 30.8
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. 29 of them aren't in the launch post.
Capabilities
- MMLU 69.6
- Global MMLU 62.7
- TruthfulQA 61.2
- HellaSwag 78.1
- HumanEval 73
- MBPP 47
- GSM8K 77.6
- MGSM 64.3
- MATH 30.8
- MathQA 33.9
- BIG-Bench Hard 64.2
- DROP 50.8
- ACPBench 62.9
- IFEval 75.2
- Multi-IF 74.7
- INCLUDE 58.2
- BLEnD 66.3
- CulturalBench 74.2
- SwitzerlandQA 67.2
- AGIEval 40.5
- ARC-Challenge 85
- GPQA Main 30.6
- GSM8K-Platinum 74.6
- MLogiQA 40.5
- RULER
Safety
- HarmBench 31.9
- RealToxicityPrompts 0.2
- LinguaSafe
Bias
- BBQ (Bias Benchmark for QA) 67.4
- ToxiGen 70.3