Gemini 4 Argon is GA: the $2/$10 price ends November 30
Gemini 4 Argon is now on the paid Gemini API at $2/$10 per million tokens until November 30, then $4/$20. Run your eval in the cheap window, not after.
Two weeks ago we told you not to rewrite your routing for a model you could not call. That changed yesterday. Gemini 4 Argon is now generally available on the paid Gemini API, in AI Studio, and for enterprises. The introductory price of $2 per million input tokens and $10 per million output has an end date now: November 30, 2026. After that, it is $4/$20. You have about seven weeks of half-price tokens to find out if Argon earns a place in your stack.
What actually happened
Per Google's announcement on October 13, Argon is open to:
- Developers on the paid Gemini API tier, through AI Studio and Android Studio
- Enterprises through the API, Gemini for business, and Google Antigravity
- Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search; Antigravity access for individuals needs Ultra
The pricing footnote matters: $2/$10 through November 30, then $4/$20 for the public model. Cached input stays 95% off. Argon also gets a 1 million token unified input-output limit through a new "continuation protocol" for long decodes, which you configure in the API. It is not on by default.
VentureBeat reports Argon at 68.90% on the Vals Index as of October 12, first place, a little ahead of Claude Sonnet 5.5 and Opus 5.5. The same report notes that Opus 5.5 still leads Argon on Google's own Terminal-Bench 4.0 numbers, 66.4% to 57.4%. Most of the other wins are Google-reported figures.
Why Gemini 4 Argon GA matters for your business
The benchmark spread between the top three frontier models is now two points. That is noise on your work. The real question is price per finished task, and you can only answer it with your own tasks.
Run the eval inside the cheap window. Take 20 to 50 real jobs from your agents: support replies, invoice extraction, a bug fix. Run them on Argon and on what you use now. Score pass rate and total tokens, not vibes. At $2/$10 the test costs half what it will in December.
Budget at $4/$20 anyway. If Argon wins only at the intro price, it does not win. Put the December number in your cost model.
Cap the long output. A 1M token limit is useful for big migrations. It is also one runaway loop away from a large bill. Set max_output_tokens on every call and only turn on the continuation protocol for the jobs that need it.
Key takeaways
- Gemini 4 Argon became generally available on the paid Gemini API on October 13, 2026
- Intro pricing of $2/$10 per million tokens ends November 30, then rises to $4/$20
- The free API tier is not a listed access channel; plan on paid usage
- Top-model benchmark gaps are about two points; test on your own tasks
- Cap output tokens and enable the 1M continuation protocol only where you need it
Seven weeks of cheap tokens is only useful if you have an eval ready to run. We build model-agnostic AI systems with eval sets, routing and spend caps you own, so testing Argon against your current model is an afternoon, not a project. See how we build it, or run your AI cost numbers at the December price.
Sources: Google, VentureBeat.
- #gemini-4-argon
- #api-pricing
- #model-evals
- #ai-costs
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Zenzap raises $44M for AI chat for frontline crews
Zenzap raised $44M to replace WhatsApp groups at work with AI chat and agents for frontline teams. If your shift schedule lives in a group text, read this.
Read itSPUR adds NYT, Economist, AFP: AI content telemetry is coming
The SPUR publisher coalition passed 80 members as NYT, The Economist and AFP joined. Its AI content telemetry standard tracks how AI reads, cites and shows your pages.
Read it