MKA Benchmark: Gemini 4 Carbon Scores 54.3% at $5.63 Per Task vs Claude's $35.28

_philschmid · x · 2026-11-17

The Multimodal Knowledge Agent benchmark from Google and Artificial Analysis spans 200 multi-step tasks across 10 verticals built around CAD blueprints, RF schematics, GIS maps, gel plates and charts. Gemini 4 Carbon scores 54.3% (Climb WSR), just behind Claude Fable 5.2's 56.6%, but at $5.63 per task versus $35.28 for Claude Fable 5.2, $10.29 for Claude Opus 5.5 and $6.15 for GPT-6 Astra.

Original post →

More from Models

Models channel →