Google's Gemini 4 Argon Surpasses GPT-6 Astra on AI Index at 60% Lower Cost

Google's AI model Gemini 4 Argon has achieved top-three status in intelligence benchmarks after nearly seven months, surpassing its previous non-Flash model. It matches GPT-6 Astra on the Artificial Analysis Intelligence Index with a score of 53 and outperforms GPT-6.1 Sol by one point. However, what's more notable is that Gemini 4 Argon achieves this at 60% of the cost per task as GPT-6 Astra, thanks to discounted pricing.

The model, developed by Google DeepMind, boasts high reasoning capabilities and has made significant strides in agentic performance. It ranks first on AutomationBench-AA with a score of 78%, outperforming Claude Sonnet 5.5 by seven points. On Terminal Bench 4, Gemini 4 Argon achieves 57%, a 53-point improvement from its previous model.

Additionally, Gemini 4 Argon has the lowest hallucination rate among leading models, with just 15% of responses containing incorrect information. However, it lags behind in accuracy, scoring 50% on AA-Omniscience compared to GPT-6 Astra's 63%. The model is currently being rolled out to selected users and features a 50% discount on pricing, which will increase to $3.98 after the promotion ends.