AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Gemini 4 Argon begins rolling out in Gemini app and Antigravity, but only if you pay

Collected Oct 13, 2026

Google started the broad rollout of Gemini 4 Argon today. The model, first shown a few weeks ago, reaches the Gemini app for general AI queries, but only on paid subscription tiers: AI Pro at $20 per month and AI Ultra starting at $99 per month. Users on lower plans and free users do not get general access. In the coding products Antigravity and AI Studio, Argon is restricted to AI Ultra subscribers. Pro subscribers can still generate code with Argon through the Gemini app's Canvas mode. Google kept earlier restrictions on its Flash models in place, leaving free users with Flash-Lite, and said it will keep Gemini 3.1 Pro available for people who need more capable models than the Flash family. Google declined to say when or whether Argon would reach Antigravity on other subscription tiers.

The model itself is unchanged from the announcement. Google continues to point to benchmarks published late last month as evidence of advanced reasoning and agentic coding capabilities. Google says the model is trained to refuse malicious instructions, and an automated misalignment monitor can halt processing if it detects an unintended workflow. Google claims those safeguards are needed because the model is so capable. Developers on the paid API can access Argon today. Pricing is temporarily $2 per million input tokens and $10 per million output tokens, with cached tokens at a 95 percent discount. That introductory rate runs only through November 30, after which it rises to $4 per million input tokens and $20 per million output tokens. Teams planning to build around Argon will need to budget for that step-up.

Google also shared new internal uses of the model. Thousands of Googlers have been using Argon. On YouTube's comment system, Argon reportedly identified redundant emoji rendering; eliminating it dropped server load by 45 percent. Engineers applied the model to the proprietary drone physics simulator used in Project Wing, where it found critical bugs that had been in the codebase since 2018. Google says Argon also advanced its robotics program, with a measurable step increase in embodied reasoning, and reports that the model is better at solving spatial problems in 3D environments.

The staged access reflects the cost of agentic coding. Multi-threaded agentic workflows in Antigravity are likely compute-intensive, which is why Pro subscribers are limited to Canvas mode rather than the full agentic environment. Google has not said whether its decision on the API price increase is tied directly to demand, but the timing suggests the company expects heavier usage after developers integrate the model. The temporary rate gives teams a window to test cost per workflow before the doubling takes effect. Cache discounts help with repeated prompts or shared context, but they apply only to cached input tokens.

For developers, the practical constraints are clear: capability comes bundled with plan-level gating. Argon is not available inside Antigravity unless the account is on AI Ultra, and API rates are scheduled to double at the start of December. Product teams evaluating Argon for agentic coding should compare it with Gemini 3.1 Pro, which Google is keeping available, and with other models they already use. The internal numbers Google cites are self-reported and cover specific systems, so they are best treated as directional. Google has not clarified whether or how the Antigravity limits will change for other subscription tiers, leaving teams to plan for the Ultra tier if they want full agentic workflows.

Why it matters: Developers and product teams who want Argon in Google's coding surfaces need an AI Ultra subscription, and API users should account for the price doubling after November 30. The clearest near-term decision is whether to prototype on Argon during the discounted window or to keep building on Gemini 3.1 Pro and the Flash family. The gating also means smaller teams on Pro plans may have to work through Canvas mode or the API rather than Antigravity's agentic workflows.

Read at Ars Technica · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Developers will get 50 percent discounted API rates through November.