MiMo V2.6 Pro at $0.2083: Vana API pay-as-you-go setup

§ What the five minutes cover

Vana API is built around one idea: one address, three protocols, minimal setup. The API address is https://api.vanaapi.com, and that address only answers API requests. Keys, balance and the usage log live on the main site, https://vanaapi.com. Nothing else needs configuring before you can send a first request.

This quick start is about MiMo V2.6 Pro, model id MiMo-V2.6-Pro. It is one of the models that speak the OpenAI format at /v1/chat/completions. On Vana API it is priced at $0.2083 per million input tokens and $0.3472 per million output tokens. Those are the numbers the rest of this page works from.

The comparison in the title is plain. Keeping your own subscription means paying a fixed amount every period whether you use it or not. Keeping an official account means managing that account yourself. Vana API is prepaid balance, charged per real token, with no monthly fee. You decide how much to put in.

§ Which client goes with which protocol

Vana API answers three protocols at the one address, and each model family has exactly one of them. MiMo V2.6 Pro belongs to the OpenAI format family, together with GPT, Grok, DeepSeek, GLM, Kimi, Qwen and MiniMax. Those models answer at /v1/chat/completions, and the clients that speak it are Cursor, Cline, Roo Code, OpenCode, Cherry Studio, any OpenAI SDK or cURL.

Claude models are different. They are only available from official Claude Code or Claude Desktop, version 2.1.285 or later, which speak Anthropic Messages at /v1/messages themselves. They do not work from any other client, any SDK, cURL or a script. Keep that boundary in mind if you also use Claude models on Vana API.

Gemini models are different again. They are only available from Gemini CLI, over the Gemini native protocol at /v1beta/models/.... They are not OpenAI-compatible. GPT models also answer on /v1/responses, which is what Codex uses. Three protocols, one address, and the client list tells you which one applies.

§ The five-minute setup

首先, open the dashboard at https://vanaapi.com and create a key. The key is what the client sends to the API address. Then open your client, for MiMo V2.6 Pro that is one of Cursor, Cline, Roo Code, OpenCode or Cherry Studio, or any OpenAI SDK or a short cURL line, and point it at https://api.vanaapi.com. Set the model to MiMo-V2.6-Pro and paste the key. That is the whole configuration.

If you prefer a script, the same OpenAI format applies. The request body names the model and the messages, the header carries the key, and the response comes back in the shape your client already expects. There is no adapter layer to write and no separate endpoint to remember. One address, one protocol for this model family, and the request goes through.

What you skip is worth naming. There is no monthly plan to choose before you start, no seat to assign and no account to provision on the model side. You top up a balance, you call the model, and the tokens you actually use are what you pay for. Everything else is left out on purpose.

§ What the numbers look like

MiMo V2.6 Pro on Vana API is $0.2083 per million input tokens and $0.3472 per million output tokens. As an example, say you run this every day with 100,000 input tokens and 20,000 output tokens. That is 3,000,000 input tokens and 600,000 output tokens over 30 days. At those rates the input side is 3 x $0.2083 = $0.6249 and the output side is 0.6 x $0.3472 = $0.20832, so $0.83322 for the month. The figure is illustrative and starts only from the two prices above.

Compare that shape with a fixed subscription. A subscription charges the same amount whether you call the model every day or once. Vana API charges per real token, so a quiet week costs less and a heavy week costs more, and the usage log in the dashboard shows each call so you can see which is which.

Other models on the same OpenAI format sit at their own prices. Qwen3.7 Plus is $0.1278 / $0.4167 and MiniMax M3 is $0.1389 / $0.5556, both per million tokens. If a task needs a different model, the address and the protocol stay the same and only the model id changes.

§ Plans, refunds and the rest

Monthly plans exist for heavier use and are separate from the per-token balance. Claude $1,000 / 30 days is $64.72 for 30 days, at most $320 a week, Claude models only. Claude $2,400 / 30 days is $153.61 for 30 days, at most $750 a week, Claude models only. GPT $4,000 / 30 days is $103.61 for 30 days, at most $1,200 a week, GPT models only. MiMo V2.6 Pro is not covered by any of them, so for this model the prepaid balance is the path.

Unused balance can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. On sign-up there is $0.10 of credit to spend on a first call. The referral programme pays 14% of what people you invite pay, 5% one level further, and earnings can be withdrawn from $13.89.

Every call is listed in the usage log in the dashboard, so the balance you see and the tokens you spent always reconcile. That is the whole surface of Vana API: one address for requests, one dashboard for keys, balance and usage, three protocols, and a price per real token. MiMo V2.6 Pro is one model id away.