MKA (AA) leaderboard
As of 16 Nov 2026, the top score on MKA (AA) is 56.6% by Claude Fable 5.2 (Anthropic), ahead of Gemini 4 Carbon at 54.3%, across 21 model configurations with a published result (Artificial Analysis, result dated 9 Nov 2026). Google DeepMind's Multimodal Knowledge Agent: 200 professional projects built on visually dense files such as drawings, schematics and scans, each graded against an expert rubric. The score is the weighted share of the harder (Climb) criteria met. Run by Artificial Analysis. MKA (AA) is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 16 Nov 2026)
- Claude Fable 5.2 holds the top MKA (AA) score at 56.6% (Artificial Analysis, 9 Nov 2026).
- Gemini 4 Carbon is second on MKA (AA) at 54.3% (Artificial Analysis, 28 Oct 2026).
- MKA (AA): Google DeepMind's Multimodal Knowledge Agent: 200 professional projects built on visually dense files such as drawings, schematics and scans, each graded against an expert rubric. The score is the weighted share of the harder (Climb) criteria met. Run by Artificial Analysis (Artificial Analysis).
- 21 model configurations have a published MKA (AA) result on BenchLeader (BenchLeader, 16 Nov 2026).
- MKA (AA) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Google DeepMind's Multimodal Knowledge Agent: 200 professional projects built on visually dense files such as drawings, schematics and scans, each graded against an expert rubric. The score is the weighted share of the harder (Climb) criteria met. Run by Artificial Analysis.
- Published by
- Artificial Analysis
- Category
- Agents & tools
- Index weight
- Reference only
- Models
- 21
- Data as of
- 16 Nov 2026
Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.
- 1Claude Fable 5.2 (max)56.6%
- 2Gemini 4 Carbon (high)54.3%
- 3Claude Opus 5.5 (max)53.0%
- 4Claude Sonnet 5.5 (max)52.0%
- 5Gemini 4 Argon (high)51.3%
- 6GPT-6 Astra (max)50.2%
- 7GPT-6.1 Sol (max)47.7%
- 8Muse Wonder (max)47.5%
- 9Kimi K3.1 Pro43.9%
- 10GPT-6.1 Terra (max)43.8%
- 11Claude Haiku 5.5 (max)41.0%
- 12Grok 4.7 (xhigh)40.1%
- 13DeepSeek V4.1 Flash 1017 (max)38.5%
- 14Qwen3.8 Max (max)37.4%
- 15MiMo-V2.6-Pro37.3%
| # | ||||
|---|---|---|---|---|
| 1 | New | 56.6% | 72.7 | 2026-11-09 |
| 2 | 54.3% | 68.9 | 2026-10-28 | |
| 3 | 53.0% | 68.4 | 2026-09-22 | |
| 4 | 52.0% | 64.2 | 2026-09-28 | |
| 5 | 51.3% | 65.3 | 2026-09-30 | |
| 6 | 50.2% | 67.3 | 2026-09-03 | |
| 7 | 47.7% | 66.6 | 2026-09-29 | |
| 8 | 47.5% | 60.8 | 2026-10-12 | |
| 9 | 43.9% | 61.8 | 2026-10-23 | |
| 10 | 43.8% | 64.1 | 2026-10-26 | |
| 11 | 41.0% | 55.3 | 2026-10-07 | |
| 12 | 40.1% | 57.4 | 2026-09-21 | |
| 13 | 38.5% | 57.7 | 2026-10-17 | |
| 14 | 37.4% | 58.7 | 2026-09-02 | |
| 15 | 37.3% | 57.5 | 2026-09-21 | |
| 16 | 36.8% | 55.1 | 2026-09-22 | |
| 17 | 36.2% | 57.8 | 2026-08-26 | |
| 18 | 35.3% | 59.4 | 2026-09-02 | |
| 19 | 35.1% | 48.3 | 2026-10-29 | |
| 20 | 33.5% | 54.3 | 2026-10-19 | |
| 21 | 27.6% | 56.0 | 2026-09-18 |
AI agent benchmark leaderboard · All LLM benchmarks
Related benchmarks
Top head-to-heads on MKA (AA)
Cite as: BenchLeader, “MKA (AA) leaderboard”, data as of 16 Nov 2026, https://www.benchleader.com/benchmarks/aa_mka.
MKA (AA): questions
- What does MKA (AA) measure?
- Google DeepMind's Multimodal Knowledge Agent: 200 professional projects built on visually dense files such as drawings, schematics and scans, each graded against an expert rubric. The score is the weighted share of the harder (Climb) criteria met. Run by Artificial Analysis. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on MKA (AA)?
- As of 16 Nov 2026, the highest score on MKA (AA) is 56.6% by Claude Fable 5.2 (result dated 9 Nov 2026), ahead of Gemini 4 Carbon at 54.3%. Scores come from Artificial Analysis and are re-read daily.
- Which AI model leads MKA (AA)?
- Claude Fable 5.2 leads MKA (AA) with 56.6% as of 16 Nov 2026 (result dated 9 Nov 2026), ahead of Gemini 4 Carbon at 54.3%.
- How many models have MKA (AA) results?
- 21 model configurations have a MKA (AA) result on BenchLeader, all taken from Artificial Analysis.
- Who runs MKA (AA) and how often is it updated?
- MKA (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does MKA (AA) count toward the BenchLeader Index?
- No. MKA (AA) is shown for reference but left out of the composite index.