Google has entered the latest round of the frontier-model race with Gemini 4 Argon, launched this week at $2 per million input tokens and $10 per million output tokens — roughly a fifth of the price charged for rival flagship models, according to launch coverage.
The Verge's coverage, summarized in Tuesday's AI news digests, notes availability is currently limited to Pro and Ultra Gemini subscribers. Google reports internal deployments lifted agentic code submissions by about 35 per cent while cutting rollbacks by a third, and says its Chrome team used the model to compress a months-long Chromium refactor into weeks.
Early community reaction has been unusually warm, with testers praising reasoning, writing and non-English cultural knowledge. The main caveat in early testing: image-to-HTML tasks were rated a tier below the strongest rival models, so the launch is not a clean sweep.
Pricing is becoming the battleground. If Argon's benchmark claims hold up in independent testing, a capable model at a fifth of rival prices would pressure every major lab to re-cut its own rates.
