AA Openness Index
How open a model actually is: weights, licence, training details and reproducibility. A measure of openness, not capability, so it is never part of the quality index.
As of 22 Sept 2026, K2 Think V2 leads AA Openness Index on BenchLeader with 88.9, ahead of K2 Think V2 at 88.9, across 36 model configurations with a published result.
- Published by
- Artificial Analysis
- Category
- Metadata
- Index weight
- Reference only
- Models
- 36
- Data as of
- 22 Sept 2026
Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.
- 1K2 Think V2 (thinking)88.9
- 2K2 Think V2 (high)88.9
- 3K2 Think V2 (medium)88.9
- 4Olmo 3.1 32B (thinking)88.9
- 5Olmo 3.1 32B88.9
- 6K2 Think V2 (low)88.9
- 7Olmo 3 7B Think (thinking)88.9
- 8Apertus 70B Instruct88.9
- 9Olmo 3 7B Think88.9
- 10Apertus 8B Instruct88.9
- 11Molmo 7B D88.9
- 12Nemotron 3 Ultra 550B A55B (thinking)83.3
- 13Nemotron 3 Super 120B A12b (thinking)83.3
- 14Nemotron Cascade 2 30B A3B83.3
- 15Nemotron 3.5 Lightning83.3
36 of 36
| # | ||||
|---|---|---|---|---|
| 1 |  K2 Think V2thinkingMBZUAIopen | 88.9 | 48.4 | 2025-12-15 |
| 2 | K2 Think V2highMBZUAIopen | 88.9 | 46.2 | 2025-12-05 |
| 3 | K2 Think V2mediumMBZUAIopen | 88.9 | 45.4 | 2025-12-05 |
| 4 |  Olmo 3.1 32BthinkingAi2open ↗ | 88.9 | 44.5 | 2025-12-12 |
| 5 | Olmo 3.1 32BAi2open ↗ | 88.9 | 43.1 | 2026-01-13 |
| 6 | K2 Think V2lowMBZUAIopen | 88.9 | 43.1 | 2025-12-05 |
| 7 | Olmo 3 7B ThinkthinkingAi2open | 88.9 | 40.1 | 2025-11-20 |
| 8 |  Apertus 70B InstructSwiss AIopen | 88.9 | 39.0 | 2025-09-02 |
| 9 | Olmo 3 7B ThinkAi2open | 88.9 | 38.8 | 2025-11-20 |
| 10 | Apertus 8B InstructSwiss AIopen | 88.9 | 37.8 | 2025-09-02 |
| 11 | Molmo 7B DAi2open | 88.9 | 34.5 | 2024-09-25 |
| 12 |  Nemotron 3 Ultra 550B A55BthinkingNVIDIAopen ↗ | 83.3 | 57.3 | 2026-06-04 |
| 13 | Nemotron 3 Super 120B A12bthinkingNVIDIAopen ↗ | 83.3 | 51.2 | 2026-03-11 |
| 14 | Nemotron Cascade 2 30B A3BNVIDIAopen ↗ | 83.3 | 50.1 | 2026-03-19 |
| 15 | Nemotron 3.5 LightningNVIDIAopen ↗ | 83.3 | 48.0 | 2026-08-11 |
| 16 | Nemotron 3 Nano 30B A3BthinkingNVIDIAopen ↗ | 83.3 | 47.0 | 2025-12-15 |
| 17 |  MiniCPM5-1BthinkingOpenBMBopen | 83.3 | 44.5 | 2026-05-25 |
| 18 | MiniCPM5-1Bno reasoningOpenBMBopen | 83.3 | 43.8 | 2026-05-25 |
| 19 | Nemotron 3 Nano 4BNVIDIAopen | 83.3 | 43.8 | 2026-03-16 |
| 20 | Nemotron 3 Nano 30B A3Bno reasoningNVIDIAopen ↗ | 83.3 | 41.7 | 2025-12-15 |
| 21 |  Granite 4.2 30BIBMopen ↗ | 72.2 | 49.7 | 2026-08-25 |
| 22 |  Intellect 3Prime Intellectopen ↗ | 72.2 | 47.1 | 2025-11-27 |
| 23 | Granite 4.2 8BIBMopen ↗ | 72.2 | 46.7 | 2026-08-25 |
| 24 | Granite 4.2 3BIBMopen ↗ | 72.2 | 45.5 | 2026-08-25 |
| 25 | Nemotron Nano 12B v2 VLthinkingNVIDIAopen | 72.2 | 41.9 | 2025-10-28 |
| 26 |  DeepSeek V4 PromaxDeepSeekopen ↗ | 44.4 | 63.7 | 2026-08-13 |
| 27 |  GLM 5.3 FlashZhipu AIopen ↗ | 44.4 | 63.6 | 2026-08-26 |
| 28 | DeepSeek V4.1 FlashmaxDeepSeekopen ↗ | 44.4 | 61.4 | 2026-09-10 |
| 29 |  Muse GlimmerhighMetaopen ↗ | 44.4 | 52.5 | 2026-08-10 |
| 30 |  Kimi K3maxMoonshot AIopen ↗ | 38.9 | 66.7 | 2026-07-16 |
| 31 |  Qwen3.8 27BxhighAlibabaopen ↗ | 38.9 | 58.9 | 2026-08-14 |
| 32 |  gpt-oss-120bhighOpenAIopen ↗ | 38.9 | 49.2 | 2025-08-05 |
| 33 | GLM 5.3maxZhipu AIopen ↗ | 33.3 | 65.4 | 2026-08-18 |
| 34 |  MiniMax M3MiniMaxopen ↗ | 33.3 | 57.2 | 2026-06-01 |
| 35 |  Mistral Medium 3.5Mistral AIopen | 33.3 | 51.1 | 2026-04-29 |
| 36 | Qwen3.8 2.4T A95BAlibabaopen ↗ | 27.8 | 64.5 | 2026-08-12 |
Cite as: BenchLeader, “AA Openness Index leaderboard”, https://www.benchleader.com/benchmarks/aa_openness_index, data as of 22 Sept 2026.
AA Openness Index: questions
- What does AA Openness Index measure?
- How open a model actually is: weights, licence, training details and reproducibility. A measure of openness, not capability, so it is never part of the quality index. Scores are reported in index; higher is better.
- Which AI model leads AA Openness Index?
- K2 Think V2 leads AA Openness Index with 88.9 as of 22 Sept 2026, ahead of K2 Think V2 at 88.9.
- How many models have AA Openness Index results?
- 36 model configurations have a AA Openness Index result on BenchLeader, all taken from Artificial Analysis.
- Who runs AA Openness Index and how often is it updated?
- AA Openness Index is published by Artificial Analysis. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does AA Openness Index count toward the BenchLeader Index?
- No. AA Openness Index is shown for reference but left out of the composite index.