Cherry Studio setup via Vana API: pay-as-you-go, $0.10 credit

§ One address, three protocols, one decision to make

Choosing an API provider for Cherry Studio usually means answering several questions at once: which address do I point the client at, which protocol does that client speak, how is the work billed, and where can I check what I have spent. Vana API is built around a simple premise that answers those questions together: one address, three protocols, minimal setup. The address for API requests is https://api.vanaapi.com. The dashboard, where keys, balance and the usage log live, sits on the main site at https://vanaapi.com. Those are two different places with two different jobs, and keeping them straight is the first practical step.

For Cherry Studio users, the relevant protocol is the OpenAI format route, served at /v1/chat/completions. Cherry Studio speaks that format, so the models it can use through Vana API are the GPT, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models listed in the price sheet. GPT models also answer on /v1/responses, which is the route Codex uses, while the other protocols on the same address serve the Anthropic Messages route and the Gemini native route. You do not need to hunt for a separate endpoint per family. The decision comes down to whether the model you want speaks the format your client already speaks.

That framing also keeps expectations honest. Vana API is a relay service, not a new client and not a new interface. You keep working inside Cherry Studio, and the change you make is a base address, a key and a model name. If your workflow already runs on the OpenAI format, the setup work is small. If you are switching clients later, the same address still covers three protocol styles, so the base configuration does not have to be rebuilt from scratch.

§ Is Cherry Studio the right client for this setup?

Cherry Studio works with Vana API through the OpenAI format at /v1/chat/completions. That means you can use any OpenAI-format model on the price list without a compatibility shim, a script or a custom wrapper. The available choices include GPT-5.6 Sol with the model id gpt-5.6-sol, GPT-6 Astra with gpt-6-astra, GPT-6.1 Sol with gpt-6.1-sol, Grok 4.7 with grok-4.7, DeepSeek V4 Flash with deepseek-v4-flash, DeepSeek V4 Pro with deepseek-v4-pro, GLM-5.3 with glm-5.3, Kimi K3 with kimi-k3, Qwen3.7 Plus with qwen3.7-plus, Qwen3.8 Max with qwen3.8-max, MiniMax M3 with MiniMax-M3 and MiMo V2.6 Pro with MiMo-V2.6-Pro.

There is one important distinction worth keeping in mind. Cherry Studio is an OpenAI-format client, so it pairs with the OpenAI-format models above. The Anthropic Messages route and the Gemini native route on the same Vana API address are for clients that speak those protocols themselves, not for Cherry Studio. In other words, the protocol is determined by the client, and the client is determined by your workflow. Cherry Studio users stay on the OpenAI format route and choose from the models listed there.

The practical setup sequence is short. Create a key in the dashboard on vanaapi.com, add credit to the prepaid balance, then set the base address in Cherry Studio to https://api.vanaapi.com and paste the key. From there, pick one of the OpenAI-format models and start sending requests. There is no monthly plan to activate before you begin, and there is no separate subscription tied to a particular model. Vana API charges per real token against the balance you have already funded.

§ What the per-token pricing looks like

Vana API prices are listed per million tokens, split into input and output. For the GPT models, the sheet also gives an official list price, which makes the difference easy to read without arithmetic tricks. GPT-5.6 Sol is $0.3472 / $1.81 at Vana API against an official list price of $4 / $20. GPT-6 Astra is $0.6944 / $3.54 against $10 / $50. GPT-6.1 Sol is $0.1389 / $0.8333 against $2 / $10. All figures are input per million tokens first, output per million tokens second.

The rest of the OpenAI-format catalogue is priced in the same per-million-token structure: Grok 4.7 at $0.4861 / $1.53, DeepSeek V4 Flash at $0.2778 / $1.04, DeepSeek V4 Pro at $1.11 / $3.33, GLM-5.3 at $0.5556 / $1.67, Kimi K3 at $1.04 / $5.21, Qwen3.7 Plus at $0.1278 / $0.4167, Qwen3.8 Max at $0.6944 / $2.08, MiniMax M3 at $0.1389 / $0.5556, and MiMo V2.6 Pro at $0.2083 / $0.3472.

Because the charge follows real token usage, the sensible question is not what a full month might cost in the abstract, but what your own traffic looks like. For example, say you send a steady stream of prompts through Cherry Studio every day. A model priced at $0.1389 / $0.8333 per million tokens will produce a different monthly figure than one priced at $1.04 / $5.21, even with identical habits, because the per-token rates and the mix of input and output differ. The usage log in the dashboard shows every call, so you can read the pattern rather than estimate it from a guess.

§ Pay-as-you-go or a monthly plan

Vana API has no monthly fee. You top up a prepaid balance, and calls are charged per real token as they happen. That model suits people who want control over spend, who are testing several models at once, or whose usage changes from week to week. Nothing is tied to a renewal date, and there is no plan to cancel when a quiet week arrives.

There is one monthly plan on the sheet, and it is narrow by design: GPT up to $2,750 / 30 days for $103.61 for 30 days, at most $820 a week, GPT models only. If your Cherry Studio workload is heavy, steady and entirely GPT-based, that plan gives a defined ceiling for the month. If you want to mix GPT with DeepSeek, Qwen, MiniMax, MiMo or the other OpenAI-format models, the pay-as-you-go balance is the route that covers the full list on one address.

A simple comparison helps here. For example, say you already know your monthly GPT usage sits comfortably under the plan limit and you will not use the other model families. The monthly plan removes the need to top up repeatedly. Say instead that you run a small daily workload and occasionally test a different model. Pay-as-you-go keeps the commitment proportional to the tokens you actually send, and the $0.10 of credit on sign-up gives you a small starting point before you add funds.

§ Money in, money out, and a record of every call

The dashboard handles the account side of the decision. Keys are created in the dashboard, the balance is visible there, and the usage log lists every call. That last point matters for pay-as-you-go work: you are not paying for a quota you cannot inspect. You can look at what was called, review the balance, and let the per-token pricing do the rest.

The refund policy is stated plainly. Unused balance can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. If you are comparing Vana API against a provider that requires a longer commitment, this window is a concrete criterion to weigh, and it is worth checking against your own top-up size before you add funds.

There is also a referral programme with figures you can read directly. You receive 14% of what the people you invite pay, and 5% one level further, with earnings that can be withdrawn from $13.89. For example, say you invite a colleague who then funds a balance; the percentage is applied to what that person actually pays, not to a list price. Treat it as a side benefit rather than a pricing argument, since the core value of Vana API for Cherry Studio remains the same: one address, three protocols, minimal setup, and billing that follows real tokens with no monthly fee.

§ Putting the decision together

Cherry Studio plus Vana API comes down to four checks. First, protocol: Cherry Studio speaks the OpenAI format, so the OpenAI-format models on the price list are the ones you can select, all through https://api.vanaapi.com. Second, address: API requests go to that endpoint, while keys, balance and the usage log live on vanaapi.com. Third, billing: prepaid balance, charged per real token, no monthly fee, with every call listed in the usage log.

Fourth, commitment. If your usage is GPT-only and predictable, the monthly plan at $103.61 for 30 days, capped at $820 a week and $2,750 / 30 days, gives you a fixed structure. If you want the wider model range or you prefer to pay only for what you use, the pay-as-you-go balance does that, and unused balance can be refunded within 3 days of the top-up under the stated terms. Either way, the configuration surface stays small: one base address, one key, a model id such as gpt-5.6-sol or deepseek-v4-pro, and a per-token price you can check in the log. That is the whole shape of the decision, and Vana API keeps it to a single address with three protocols behind it.