What changed: Google, together with Artificial Analysis, announced the Multimodal Knowledge Agent (MKA) benchmark, which measures how well AI agents reason across visually complex files such as blueprints, schematics, decks, and spreadsheets. The benchmark comprises 200 tasks built with experts across 10 verticals and 83 occupations. In the evaluation, Gemini 4 Carbon ranked #2 in quality and completed tasks at 6x lower average cost than the top-scoring model.
Who is affected
According to the source, this covers businesses across 10 verticals and 83 occupations, including architecture, biology, underwriting, and data science, that use the Gemini API or similar AI agents on visual documents.
What it means for your business
For businesses using the Gemini API, this means a lower-cost model option, Gemini 4 Carbon, that performs strongly on visual document tasks like schematics, reports, and spreadsheets.
What to watch
No action is required right now. If you use the Gemini API, evaluate Gemini 4 Carbon's cost-to-performance results for your own visual document tasks via Google AI Studio.
Timing
The source states this took effect 'today'; the announcement and Gemini 4 Carbon's availability via the Gemini API/Google AI Studio apply as of the November 10, 2026 publication date.
