The short answer
Whizi carries Google Gemini models inside the same subscription as GPT, Claude and the rest of a 280+ model catalogue, so no separate Google AI subscription is needed. One Gemini Flash row is on Starter at 1 credit per message, seven further Google identifiers are on Pro, and everything else in the family resolves to Powerhouse.
What a Gemini message costs, at a glance:
| Plan | Gemini rows it opens | Credits per message | Monthly credits |
|---|---|---|---|
| Free | Any text model, inside a 7 message lifetime allowance | not metered in credits | 0 |
| Starter | Gemini 3 Flash | 1 | 400 |
| Pro | Seven Google identifiers, listed below | 1 to 10 | 2,000 |
| Powerhouse | Every remaining Gemini identifier in the catalogue | varies by row | 8,000 |
Starter is $15.99/month, or $10.99/mo billed annually at $131.88. Pro is $29.99/month, or $19.99/mo billed annually at $239.88. Powerhouse is $49.99/month, or $34.99/mo billed annually at $419.88. New accounts start with a 7 day trial for $0.99.
The priced Gemini rows, with credit cost and plan
The Gemini rows below are the ones this page covers, ordered cheapest first. The credit column is what a single message on that row spends from your monthly allowance. The last two columns come from the Whizi Cost Index, which prices a "standard answer" of 1,000 input tokens plus 500 output tokens so that list rates from different providers become comparable. Those prices were read from OpenRouter on 20 August 2026.
| Model | Credits per message | Plan required | Context window | Cost per standard answer |
|---|---|---|---|---|
Gemini 3 Flash (gemini-3-flash-preview) | 1 | Starter | not in the price index | not in the price index |
| Gemini 2.5 Flash Lite | 1 | Pro | 1M | $0.0003 |
| Gemini 3.1 Flash Lite | 1 | Powerhouse | 1M | $0.001 |
| Gemini 3.7 Flash | 2 | Pro | 1M | $0.001313 |
| Gemini 2.5 Flash | 2 | Pro | 1M | $0.00155 |
| Gemini 3.5 Flash Lite | 2 | Powerhouse | 1M | $0.00155 |
| Gemini 3.6 Flash | 3 | Pro | 1M | $0.002625 |
| Gemini 3.5 Flash | 8 | Pro | 1M | $0.006 |
| Gemini 2.5 Pro | 8 | Pro | not in the price index | not in the price index |
| Gemini 3.1 Pro Preview | 10 | Pro | 1M | $0.008 |
| Gemini 3 Pro Image Preview | 10 | Powerhouse | not in the price index | not in the price index |
One caveat on how to read that table. The Cost Index prices 100 rows for comparison, not the whole catalogue, so a blank price cell means the row was not in that sample and not that the model is free. How the credit figure itself is arrived at is set out in the credits reference.
One row is priced by owner policy rather than by arithmetic. The Starter Gemini Flash identifier is charged 1 credit against a cost-true rung of 2, which is what allows a 400 credit Starter allowance to be 400 Gemini messages rather than 200.
The median row in the price index answers for $0.00185. Most of the Gemini Flash rows above sit at or under that figure, and the one priced Pro-class Gemini row answers for $0.008 against $0.105 for the priciest row in the index.
What each plan opens, by identifier
The gate reads the model identifier and returns a minimum tier, which is why the Starter and Pro sides below can be written out identifier by identifier at all. The models reference covers the enforcement model; what follows is only the Google half of it.
Starter reaches four identifiers in total, one of which is Google: auto, whizi-ai, openai gpt-5.6-luna, and google gemini-3-flash-preview. So on Starter, Google is exactly one fast Gemini row and nothing else.
Pro folds in the whole Starter set and adds 37 further identifiers, of which seven are Google: gemini-3.7-flash, gemini-3.6-flash, gemini-3.5-flash, gemini-3.1-pro-preview, gemini-2.5-pro, gemini-2.5-flash, and gemini-2.5-flash-lite. That is the current Flash tier, the Flash Lite row, and the Pro-class rows.
Powerhouse is the fallback. Any identifier that is not in the Starter set, not an agent persona, and not in the explicit Pro list resolves to Powerhouse, and that includes every Gemini identifier the catalogue gains until it is deliberately promoted into the Pro list. That is the safe direction rather than an oversight.
If you select a Gemini row above your tier, the request is refused server-side with HTTP 403, code tier_upgrade_required, and the message "Upgrade your plan to use this model." See why a model shows as unavailable.
Using Gemini without a Google AI subscription
Whizi pays the providers and bills you once, so a Whizi plan is the only subscription involved. There is no Google account to connect and no Google AI key to paste.
Before paying anything, a free Whizi account gets a lifetime allowance of 7 messages that never resets. That allowance is deliberately not model-gated: the funnel lets prospects try any text model inside the cap, so the Gemini rows are reachable on a free account even though the free credit limit and free message limit are both 0. When the allowance is gone the reply is HTTP 402, code free_limit_reached, with the message "Your free messages are used up. Start a subscription to keep chatting."
After that, what a Gemini allowance actually buys:
| Plan | Monthly credits | Messages on the 1 credit Gemini row | Messages on a 2 credit Gemini row | Messages on the 10 credit Gemini Pro row |
|---|---|---|---|---|
| Starter | 400 | 400 | not included | not included |
| Pro | 2,000 | 2,000 | 1,000 | 200 |
| Powerhouse | 8,000 | 8,000 | 4,000 | 800 |
Credits do not roll over: the balance sums only the rows carrying the current period key, so last period's unspent credits are unreachable. The credits reference has the mechanism.
Where Gemini runs without you picking it
Auto routing. Two of Auto's six rungs are Gemini Flash rows. Any turn carrying an attachment is routed to Gemini Flash at 2 credits, on the reasoning that an image needs vision rather than a large context window, and the rewrite rung is the same Gemini Flash row at 2 credits. Auto is a router, not a model, so it carries no rung of its own: a turn costs whatever the picked model costs. Starter cannot reach that 2 credit Gemini Flash row, so a Starter account that sends an attachment is answered on the house model at 1 credit instead. The per-tier ranges are free 1 to 1, Starter 1 to 1, Pro 1 to 10, Powerhouse 1 to 20.
Voice mode. Whizi voice runs on the Gemini Live API. The worker mints a short-lived session token and the browser opens a WebSocket straight to Google, so no audio passes through Whizi's own backend. Sessions are capped at 15 minutes and the token is minted with a 16 minute lifetime, one minute of slack past the cap. Replies use a single warm preset voice. Voice is metered in minutes rather than credits: Starter 10 minutes per month, Pro 80, Powerhouse 500, and free accounts 0. The weekly allowance is the monthly figure divided by four and rounded up, which gives Starter 3 minutes, Pro 20 and Powerhouse 125. See voice mode will not start.
Two Gemini facts inside the picker
The Recommended row. The top section of the model picker is the house model, Auto, and one current row each from GPT, Claude and Gemini. That section is constrained to models reachable at Pro or below, so the top of the picker is never Powerhouse-locked.
Image generation is separate. The Google rows in the Image Generation section of the picker carry no per-message credit rung: they are priced in the Cost Index but absent from the credit table, as are the other standalone image generators in that section. The Gemini image preview row in the model table above is not one of them and does carry a rung, at 10 credits. Media generators skip the per-model gate and are tiered by picker section instead, against a monthly image count: free 0, Starter 0, Pro 100, Powerhouse 500. Generated media is retained for 30 days.
What a 1M context window does and does not buy
Almost every Gemini row in the table above carries a 1,000,000 token context window, and that is the provider's figure, not the per-turn budget Whizi sends. What the large window buys on a Gemini row is that it clears the 93,000 token line by a wide margin, so it always gets the flat per-turn budget rather than the smaller proportional one Whizi applies to short-window models. It does not mean a single message can carry a million tokens. The budget itself is set out in when a conversation is too long.
Two details that change the arithmetic. OpenRouter counts input plus requested maximum output against a model's window, not input alone. And Whizi's token estimator can undercount a real tokenizer by up to 1.66x on JSON, so an allowance of 1.8x is applied to input before a request goes out. Both push in the conservative direction.
Web search is a per-request toggle rather than a per-model capability flag. How the toggle works is covered in the web search reference, and switching models mid-conversation covers moving a thread from Gemini to another family.
- A Whizi plan is the only subscription needed, with no Google account to connect
- Starter opens one Gemini Flash row at 1 credit per message
- Pro is the tier that reaches the Pro-class Gemini rows
- Every other Gemini identifier resolves to Powerhouse by default
- Auto routes any turn carrying an attachment to a Gemini Flash row at 2 credits
- Voice mode runs on the Gemini Live API and is metered in minutes, not credits
- The Google rows in the Image Generation section carry no credit rung and draw on the monthly image count
- A free account can try any text model inside a 7 message lifetime allowance
Frequently asked questions
Can I use Gemini without a Google AI subscription?
Yes. A Whizi subscription is the only plan involved. Starter at $15.99/month opens a Gemini Flash row at 1 credit per message, and Pro adds the Flash tier, the Flash Lite row and the Pro-class Gemini rows.
How many credits does a Gemini message cost in Whizi?
From 1 to 10, depending on the row. The Starter Gemini Flash identifier is 1 credit, the Flash and Flash Lite rows run from 1 to 8, and the Pro-class rows sit at 8 and 10. The per-row figures are in the table above.
Which Whizi plan do I need for Gemini Pro?
Pro. The Pro-class Gemini rows sit in Pro's Google set, alongside the Flash tier and the Flash Lite row, and the exact identifiers are listed above. Starter reaches one Gemini Flash row and no Pro-class Gemini model, so a Starter account that selects one gets HTTP 403 with the message "Upgrade your plan to use this model."
Do I get the full 1 million token context window on Gemini in Whizi?
No. The window is the provider's figure, and it is 1,000,000 tokens on almost every Gemini row in the table above, but what Whizi sends per turn is its own budget: a flat 40,000 input tokens and 20,000 output tokens. What happens once a thread outgrows that is in when a conversation is too long.
Is Whizi voice mode running on Gemini?
Yes. Voice runs on the Gemini Live API, metered in minutes rather than credits: 10 per month on Starter, 80 on Pro, 500 on Powerhouse, and 0 on a free account.
Does Whizi list which Gemini models can read images or browse the web?
No, and there is no list to publish, because the catalogue carries no per-model vision or search capability field. Web search is a per-request toggle rather than a per-model flag, and per-model knowledge cutoff dates are not recorded either. Auto routing settles the vision question in one direction on its own: any turn carrying an attachment is routed to a Gemini Flash row.