Gemini 4 Argon Ties GPT-6 Astra at 53 With 15% Hallucination Rate

Google DeepMind's Gemini 4 Argon ties GPT-6 Astra at 53 on Artificial Analysis's Intelligence Index with high reasoning enabled, one point above GPT-6.1 Sol. That is a 23-point jump over Gemini 3.1 Pro Preview.

The hallucination figure is the one that stands out. Artificial Analysis calls Argon's result "a 15% hallucination rate, the lowest of any model scoring 45+ on the Intelligence Index." GPT-6 Astra sits at 51% on the same measure; GPT-6.1 Sol at 54%.

Argon ranks first on AutomationBench-AA at 78%, posts 57% on Terminal Bench 4, scores 50% on AA-Omniscience, and lands an AA-Briefcase Elo of 1494.

Pricing is the other unusual beat. The current rate is $2 input and $10 output per million tokens, which the firm flags as a 50% promotional discount; cached inputs are $0.10 per million. Per Intelligence Index task that works out to $1.99 against $3.98 at standard pricing, roughly 60% of what GPT-6 Astra costs for equivalent intelligence. Google has not confirmed how long the promo runs.

The model is rolling out to selected users only, with a 1M-token context window and support for text, image, video and speech input. Artificial Analysis frames the release as "Google's first proprietary model above the Flash class in over 7 months," a line that reads differently alongside staff accounts that Gemini 4 aces benchmarks but stumbles on coding across the 198 Google stories we have tracked in the past 90 days.