Gemini 4 Argon (High): Intelligence, Performance and Price Analysis

Commenters argued about whether Gemini 4 Argon is a meaningful step forward or just more of the same. Some praised its introductory pricing ($2/1M input, $10/1M output, heavy cached-token discount) and low hallucination rate, saying it competes with Opus 5.5 (high) and Astra on cost-per-task and efficiency, and noting Google’s token-conservative behavior. Others called it underwhelming, saying it does not improve intelligence or price compared to peers, that benchmarks show it trailing or matching rather than beating frontier models, and that delayed launch for “safety reasons” looked comical. Several commenters pointed out nuance about benchmark tiers (high vs max) and warned that direct comparisons may be unfair until an “xhigh”/Ultra variant or further tuned checkpoints appear.

Discussion also touched on strategy and market dynamics. Some commenters said Google is unlikely to start a price war and instead leverages vertical integration and search-driven token volume for long-term advantage, while others argued Google is losing market share and will have to buy relevance. A few noted the practical limits of evaluation because Google hasn’t made the model public, and disputed claims about competitors’ revenue or “unlimited” usage, with back-and-forth over what plan economics actually mean for heavy users.