TL;DR
Google has made Gemini 4 Argon generally available on Wednesday 14 October: through the Gemini API for developers and enterprises; in Google Antigravity for AI Ultra subscribers; and, for Ultra and Pro subscribers, inside the Gemini app and Search’s AI Mode. Input is priced at $2 per million tokens and output at $10, with cached input discounted by 95%. Every performance claim in the launch post is Google’s own measurement, and the one UK data point comes from elsewhere: the AI minister said on Monday that AISI scientists were already evaluating Gemini before release.
What Google claims
Google says Argon tops the Vals Index, a benchmark that scores models on tax, legal, finance and coding work, and that it handles a unified 1 million token input-output limit through a new continuation protocol. Internally, it credits Argon in Antigravity with a rise of over 35% in agentic code submissions and almost a third fewer rollbacks, and cites a Chrome refactor across a six-million-line legacy folder, a 40% cut in session preparation time for its open-source Agent Development Kit, and a 2.85x inference speed-up for its Gemma 4 E2B model on an A100.
Science examples run from quantum error correction, where a translated AlphaQubit kernel nearly doubled throughput, to enzyme design and new counterexamples to Grad’s conjecture in fusion research. Google is offering 10,000 scientists six months of Antigravity’s top usage tier free.
Cyber: wider access, fewer guardrails
Google is extending Argon to hundreds more partners in its Fairwind Program, which gives vetted defenders access without cyber guardrails. Security firm Wiz, an early user, says the model helped find a bypass of a public file download check in healthcare software, which had left a master key exposed and granted administrator access. On the defensive side, Google reports that external bio red-teaming unlocked less than 1% of the sub-tasks it tests, that shortcut attempts on coding benchmarks fell to 2%, and that attack success on Gray Swan’s indirect prompt injection test was cut to 0.7%.
The UK reading
Resultsense covered Google’s initial Argon announcement earlier this month. General availability opens the model to UK developers and enterprises at the same dollar pricing as everywhere else. The safety claims will not stand alone for long: Kanishka Narayan told an Imperial College audience on Monday that AISI’s scientists were “conducting rigorous evaluations of the latest generations of models”, Gemini among them, after Washington lifted its block on pre-release access on Sunday. In our view that makes Argon an early test of whether a UK evaluation can be published alongside a vendor’s model card.
Looking forward
Google calls Argon “just the beginning”, with more Gemini 4 models to follow, and the introductory price carries a footnote, so buyers should expect it to move. Whether AISI publishes its findings on Argon is the question for UK readers.