is gemini 4 argon available yet?

as of october 2026, yes, if you pay. google opened gemini 4 argon on october 13, 2026, with a launch post that says it is "giving developers, enterprises, scientists, and consumers access" to the model.

so is gemini 4 argon available to you? in practice, the developers, enterprises and consumers on that list are paying customers: the paid gemini api, google ai pro and ultra subscribers in the gemini app, and ultra subscribers in google antigravity. scientists and vetted defenders can also apply for program access. free gemini, google ai plus and the free api tier still do not get it.

our verdict: if you already pay for google ai pro, argon is rolling out to your gemini app now. if you build on the api, test it before the introductory price ends on november 30, 2026. keep claude opus 5.5 for terminal heavy agent work, where google's own chart still has opus ahead.

what changed on october 13?

argon went from a gated release to a paid product. when we covered the september 30 announcement, access was gated to groups such as googlers, a set of fairwind program partners and early access testers. google said the wider release would start with paid api customers and ultra subscribers, with no date.

google's october 13 launch post opens it to developers and enterprises through the api, and to pro as well as ultra subscribers in the gemini app and ai mode in search. pro was not named in the september plan. google says argon is rolling out globally. the other changes are smaller, but they matter if you fit one of the groups:

• fairwind grows. google says it is beginning to expand argon to hundreds of additional fairwind program partners. defenders can apply for a version without cyber guardrails, on a rolling basis.
• scientists can apply. google is offering 10,000 scientists the highest antigravity usage tier free for 6 months, onboarded on a rolling basis and subject to eligibility, capacity and selection.
• developers get a stable model. google's model page lists argon as stable, with a 1,048,576 token window shared by input and output. the default synchronous output cap is 65,536 tokens, and long decode raises it to the full window.

most frontier models in our context window comparison ship a window of roughly that size. what stands out is how much of it can go to output. google calls the unified limit industry leading.

which gemini plans include gemini 4 argon?

google ai pro is the cheapest way in. google's us plans page now tells visitors to pick ai pro or ai ultra to get argon, and lists it on neither free gemini nor google ai plus.

businesses have their own route. google says enterprises can use argon in gemini for business, the api and antigravity.

there is one catch for developers on a pro plan. google's api pricing page says argon is not available on the free tier or on the google ai pro plan. in our reading, pro buys you argon in the gemini app and ai mode, not in ai studio, so developer access means ultra or a paid api key. the pro plan does list entry antigravity limits, but the plans page does not name the model behind them.

if you are on free gemini, this launch does not change your week. our chatgpt free vs gemini free comparison still describes what you get.

how much does the gemini 4 argon api cost?

argon launched at an introductory price that ends on november 30, 2026, so the gemini 4 argon price you see today is not the one you will pay in december. google's api pricing page lists these standard paid tier rates, per million tokens:

• through november 30: $2 input, $10 output, $0.10 for cached input
• from december 1, 2026: $4 input, $20 output, $0.20 for cached input
• caching: implicit only, since google says explicit caching is not supported

the december rate is the same list price anthropic charges for claude opus 5.5, per anthropic's pricing docs. what you actually pay also depends on how many tokens each model burns.

on the vals index, which prices both models at that same list rate, vals lists a cost per test of $15.68 for argon and $32.14 for opus 5.5. argon's run also used fewer input and output tokens. treat that as one benchmark's workload, not your invoice. our ai model pricing comparison covers the rest of the market.

is gemini 4 argon better than claude opus 5.5?

on google's own chart, argon leads every knowledge work row and trails opus on terminal-bench 4.0. the figures below come from the deepmind gemini page. google ran argon at its highest thinking setting unless noted, and used the maximum reported settings for opus where available.

read the coding rows with care. google's methodology notes say argon's deepswe and terminal-bench scores are google's own runs. the opus terminal-bench figure matches the one in our best ai model for coding guide.

the third party check is vals. as of october 13, 2026, the vals index ranks argon first at 68.90%, ahead of claude sonnet 5.5 at 67.04% and opus 5.5 at 66.97%.

vals adds a footnote that cuts in argon's favor. some opus, sonnet and fable tasks were served by a fallback model after a provider refusal. counted as failures, opus drops to 65.05%, and argon had no fallbacks. in our view, that makes argon's lead real but narrow.

should you switch from claude opus 5.5 to argon?

switch some work, not all of it. in our view, the gemini 4 argon vs claude opus 5.5 choice comes down to what your agents spend their day doing. here is how we would split it as of october 15, 2026:

• try argon for long repo tasks, where google's chart has it ahead on deepswe v1.1.
• try argon for finance agent work, where vals ranks it first on finance agent v2.
• try argon for very long inputs. graphwalks at 256k to 1m tokens is argon's widest margin over opus in google's chart, in percentage points.
• keep opus 5.5 in claude code for terminal heavy agents, since google's own chart still has opus ahead on terminal-bench 4.0.
• test before december 1, while argon's api price per token is below opus's.
• on google ai pro, pick argon in the gemini app. you already pay for it.

for the wider ranking, see claude vs chatgpt vs gemini. for meta's october 12 launch against the same opponent, see muse wonder vs claude opus 5.5. in our view, expect this ranking to move again before the introductory price expires.

keep your context when the model changes

our take: you will probably switch models again before the year ends, and your working context should not live inside any one vendor's app.

remynd is a mac app that keeps that record for you. it captures the focused window, runs ocr locally with apple vision, and keeps the searchable index on your mac. you can exclude apps and sites, and recordings are kept for 30 days by default.

its agent window runs your own claude code or codex cli against that history.

download remynd for mac and keep your memory when you change models.