Google has fired its biggest shot yet in the global race for artificial intelligence, releasing its flagship Gemini 4 Argon model to developers and businesses worldwide and pricing it sharply below the systems of OpenAI and Anthropic.
The launch pitches Argon as a “deep work” system for complex, multi-step tasks across coding, finance, law, cybersecurity and science. Introductory API pricing starts at $2 per million input tokens and $10 per million output tokens, with rates set to double after November 30 — a level below Anthropic’s Claude Opus 5.5 ($4 and $20) and OpenAI’s GPT-6 Astra ($10 and $50).
Estimates cited in coverage put Argon’s cost per benchmark task at roughly $1.99, about three-fifths of what GPT-6 Astra costs, though such comparisons vary with the task. For enterprises deploying AI agents at scale, the pricing is a direct challenge to the idea that the most capable models must also be the most expensive — as Washington tightens export rules on the advanced chips that power systems like these.
A comeback after months of setbacks
The release follows a bruising period for Google’s AI ambitions. The company confirmed in September that it no longer planned to release Gemini 3.5 Pro, and it has overhauled its DeepMind laboratory after co-founder Demis Hassabis stepped aside as chief executive and several leaders of the Gemini model team departed. Argon is the company’s first major frontier release since Gemini 3 arrived in late 2025.
Google says teams using Argon inside its Antigravity coding tool raised agentic code submissions by more than 35 per cent while cutting rollbacks by nearly a third. On one industry leaderboard, the model placed just behind Claude Opus 5.5 and Claude Sonnet 5.5. Early testers include Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs — and the strain that AI workloads place on data-centre power and water use is drawing fresh scrutiny of its own.
Three labs, one cost war
Commentators say the timing matters as much as the technology. Argon arrived a day after Meta’s own frontier launch, and analysts argue that back-to-back releases from Google and Meta have broken what briefly looked like an OpenAI-Anthropic duopoly, turning the contest into a three-way fight on price as well as capability.
Paid subscribers can use Argon through the Gemini app and AI Mode in Search, developers through the Gemini API and the company’s AI-building tools, and Google says it will open its Fairwind programme to hundreds more partners, giving vetted cybersecurity defenders a version of the model without its cyber guardrails. The company is also offering 10,000 scientists six months of free access to its highest-usage coding tier.
