Claude Fable 5.2 leads DeepMind's new multimodal benchmark
Google DeepMind's new Multimodal Knowledge Agent benchmark ranks Claude Fable 5.2 first with a 57% score, but it costs about five times more per task than the runner-up, Gemini 4 Carbon, at 54%.
Artificial Analysis has launched Google DeepMind's Multimodal Knowledge Agent (MKA) evaluation, which is available only on its platform. The benchmark tests agents on knowledge work that depends on complex visual files.
MKA has 200 private tasks drawn from real work across 10 industries. Agents have to work from dense inputs such as architectural drawings, engineering schematics and clinical scans. They then produce deliverables such as slide decks, spreadsheets, PDFs, Word documents and marked-up images. Artificial Analysis says the visual inputs are needed to complete each task rather than serving as extra context.
Each task is graded against expert rubrics that average about 70 weighted criteria. These are split into "Regression" checks, which are baseline checks, and harder "Climb" checks. The headline metric is Climb Weighted Satisfaction Rate (WSR). Tasks run in Stirrup, Artificial Analysis's open-source agent harness, for up to 250 turns, with code execution, web search, web fetch and image-viewing tools. GPT-6 Luna (medium) grades the output as an agentic judge.
Claude Fable 5.2 (max) scored 57%, ahead of Gemini 4 Carbon (high) at 54% and Claude Opus 5.5 (max) at 53%. Costs differ widely. Fable 5.2 averaged $31.97 per task, compared with $6.29 for Gemini 4 Carbon and $1.11 for GPT-6.1 Sol. All four OpenAI models tested were on the cost-performance Pareto frontier below Carbon.
No model led all 10 industries. Fable 5.2 led five, and Gemini 4 Carbon led Healthcare. Word and text-file outputs were the hardest deliverables, and no model scored above 52% on either.