Google and Artificial Analysis introduce MKA
November 10 benchmark targets visual document work.
MKA, co-developed by Google and Artificial Analysis, has 200 private evaluation tasks and 10 separate public examples.
This co-developed benchmark is not independent validation or a regional product rollout. Hong Kong, Macau and Taiwan eligibility, clients, Traditional/Simplified Chinese, Cantonese and pricing remain unverified. Tests were not reproduced; scores do not establish performance across all work.
This co-developed benchmark is not independent validation or a regional product rollout. Hong Kong, Macau and Taiwan eligibility, clients, Traditional/Simplified Chinese, Cantonese and pricing remain unverified. Tests were not reproduced; scores do not establish performance across all work.
Usage conditions
- Hong Kong
- Local access conditions unverified; tests not reproduced.
- Macau
- Local access conditions unverified; tests not reproduced.
- Taiwan
- Local access conditions unverified; tests not reproduced.
Official source: Introducing the Multimodal Knowledge Agent benchmark