AI Rate CardGemini API pricing and cost: Gemini 4 Argon $2 / $10, Gemini 3.8 Flash $0.75 / $3.75 per 1M tokens
Updated Nov 2, 2026
Google sells two current Gemini models most people choose between: Gemini 4 Argon, the large model, and Gemini 3.8 Flash, the fast and cheaper one. Both are resold by more than twenty providers - Google's own Vertex AI, routers, aggregators and developer tools. Most of them charge exactly Google's list price, a few charge more for a region or a second tier, and a few charge less. The tables below are rebuilt from today's prices on every page view.
The two models at a glance
In the 15 October 2026 collection the typical (median) price per 1 million tokens, input / output, was:
- Gemini 4 Argon - $2 / $10, cached input $0.10. 20 providers sell it.
- Gemini 3.8 Flash - $0.75 / $3.75, cached input $0.075. 26 providers sell it. Both take up to 1,048,576 tokens of context. Argon costs about 2.7 times as much as Flash on both input and output. Side by side: compare the two.
Gemini 4 Argon: every provider
| Provider | Model | Input, $ | Output, $ |
|---|---|---|---|
| openrouter | Google: Gemini 4 Argon (batch) | $1.00 | $5.00 |
| cometapi | Gemini 4 Argon | $1.60 | $8.00 |
| iteracompute | Gemini 4 Argon | $1.60 | $8.00 |
| Gemini 4 Argon | $2.00 | $10.00 | |
| abacus | Gemini 4 Argon | $2.00 | $10.00 |
| aihubmix | Gemini 4 Argon | $2.00 | $10.00 |
| aiml | Gemini 4 Argon | $2.00 | $10.00 |
| edenai | Gemini 4 Argon | $2.00 | $10.00 |
| edenai | Gemini 4 Argon | $2.00 | $10.00 |
| edenai | Gemini 4 Argon | $2.00 | $10.00 |
| edenai | Gemini 4 Argon | $2.00 | $10.00 |
| fastrouter | Google: Gemini 4 Argon | $2.00 | $10.00 |
| github-copilot | Gemini 4 Argon | $2.00 | $10.00 |
| impossibl | Gemini 4 Argon | $2.00 | $10.00 |
| jiekou | gemini-4-argon | $2.00 | $10.00 |
| kilo | Google: Gemini 4 Argon (50% off) | $2.00 | $10.00 |
| meganova | Gemini-4-Argon | $2.00 | $10.00 |
| openrouter | Google: Gemini 4 Argon | $2.00 | $10.00 |
| perplexity-agent | Gemini 4 Argon | $2.00 | $10.00 |
| requesty | Gemini 4 Argon | $2.00 | $10.00 |
Live data: this table is built from today's prices.
What the table shows on 15 October:
- The list price is the norm. Google AI Studio, GitHub Copilot, Vercel, Kilo, FastRouter, OfoX and a dozen more charge $2 / $10.
- A few undercut it. CometAPI listed $1.60 / $8, 20 % below Google.
- Regions cost more. Requesty's EU endpoint is $2.20 / $11 (+10 %).
- A second, higher figure. Vertex AI, ZenMux and OpenCode list $4 / $20 - twice the list price. Before you pick by the lowest number, open the card: the higher figure may belong to long prompts or a priority tier, and the conditions are on the provider's page.
- No free tier. Google gives no free quota for Argon: every request is paid.
Gemini 3.8 Flash: every provider
| Provider | Model | Input, $ | Output, $ |
|---|---|---|---|
| kenari | Gemini 3.8 Flash | $0.33 | $1.65 |
| openrouter | Google: Gemini 3.8 Flash (batch) | $0.38 | $1.88 |
| cometapi | Gemini 3.8 Flash | $0.60 | $3.00 |
| iteracompute | Gemini 3.8 Flash | $0.60 | $3.00 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| Gemini 3.8 Flash | $0.75 | $3.75 | |
| 302ai | Gemini 3.8 Flash | $0.75 | $3.75 |
| abacus | Gemini 3.8 Flash | $0.75 | $3.75 |
| aihubmix | Gemini 3.8 Flash | $0.75 | $3.75 |
| aiml | Gemini 3.8 Flash | $0.75 | $3.75 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| edenai | Gemini 3.8 Flash | $0.75 | $3.75 |
| fastrouter | Google: Gemini 3.8 Flash | $0.75 | $3.75 |
| github-copilot | Gemini 3.8 Flash | $0.75 | $3.75 |
| impossibl | Gemini 3.8 Flash | $0.75 | $3.75 |
| jiekou | gemini-3.8-flash | $0.75 | $3.75 |
| kilo | Google: Gemini 3.8 Flash (50% off) | $0.75 | $3.75 |
Live data: this table is built from today's prices.
What the table shows on 15 October:
- $0.75 / $3.75 at most providers, Google included.
- Cheaper resellers exist. CometAPI listed $0.60 / $3, Kenari about $0.33 / $1.67. One router (routing.run) listed $0.04 / $0.21 - 95 % below Google; read its terms before you rely on a price that low.
- Regions and tiers cost more. Requesty EU $0.825 / $4.125; Vertex AI, ZenMux and OpenCode $1.50 / $7.50.
The free tier: free, but you pay with your prompts
Gemini 3.8 Flash can be used without paying in Google AI Studio. The catch is written in Google's own terms: on the free tier your prompts and answers may be used to improve Google's products, and people may read them. On the paid tier they are not. Our catalog records this condition model by model (free tier available, "used to improve products" yes or no), so it shows on the card next to the price.
The free tier is also limited: requests per minute and per day are capped, and the caps change without notice. We are starting to collect those limits every day as well; until then, check Google's rate-limit page before you plan a product on the free tier.
Rule of thumb: the free tier is for trying a model, not for customer data. For anything private, use a paid key - at Google or at any provider in the table.
How to choose
- Most tasks - chat, extraction, summaries, agents that call tools: start with Gemini 3.8 Flash. It is 2.7 times cheaper and fast.
- Hard reasoning, long documents, code you cannot afford to get wrong: Gemini 4 Argon.
- Price per provider is mostly the same for both models. Pick a provider for what else it gives you - one bill for many models, a region you need, a limit high enough for your load - and check the card for the conditions behind a price that looks too good.
- To estimate a month, put your own volume into the cost calculator: it ranks every offer of a model by your monthly bill.
All prices are in USD per 1 million tokens, without taxes, as the providers publish them; each row links to its card with the source and the date it was checked.