Google's Gemini 4 Carbon tops three agent benchmarks at under half the cost per task

Google released Gemini 4 Carbon to developers, enterprises, scientists, and consumers on Tuesday. The company says its latest frontier model is the top performer on three independent agent benchmarks at less than half the per-task cost of the next-best outside model. Gemini 4 model leads CursorBench 4.0, WANDR and AutomationBench In the launch post , Google DeepMind SVP Koray Kavukcuoglu wrote that Carbon topped CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench. Google says it costs less than half what the next closest outside model costs per task. Google built Carbon for long, multi-step jobs. Complex coding and debugging, open-ended scientific research, and cybersecurity are its targets. Separately, Google cooperated with Artificial Analysis to build the Multimodal Knowledge Agent benchmark.

BeCrypto News shows a short excerpt and links to the publisher. The full report belongs to Cryptopolitan.