GPT-6.1 Sol on Vana API: one address, $0.1389 per 1M, pay-as-you-go

§ One address, three protocols, minimal setup

A student project usually begins with a small decision: where do the model requests go? With Vana API, that decision collapses into one address, https://api.vanaapi.com, and three protocols behind it. You point your client at that single endpoint, and the protocol your client speaks decides the route. There is no second gateway to configure and no separate integration to maintain. The setup is deliberately small, because an API relay should get out of the way of the work you are actually trying to do.

Here is the whole map. GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax, and MiMo models use the OpenAI format (/v1/chat/completions), and they answer from Cursor, Cline, Roo Code, OpenCode, Cherry Studio, any OpenAI SDK, or cURL. GPT models also answer on /v1/responses, which is what Codex uses. That is the route GPT-6.1 Sol takes in an OpenAI-style client.

The other two protocols are narrower, and it helps to know them before you start. Claude models work only from official Claude Code or Claude Desktop, version 2.1.285 or later, because those clients speak Anthropic Messages (/v1/messages) themselves. They do not work from any other client, any SDK, cURL, or a script. Gemini models work only from Gemini CLI, over the Gemini native protocol (/v1beta/models/...). They are not OpenAI-compatible.

One address, three protocols, minimal setup. The dashboard for keys, balance, and the usage log sits on the main site, https://vanaapi.com, while https://api.vanaapi.com answers API requests and nothing else. You sign in once on the main site, create a key there, and paste that key into the client you already use.

§ A five-minute setup for GPT-6.1 Sol

Start on https://vanaapi.com, create an account, and open the dashboard. Keys are created in the dashboard, and a key is the only credential a quick start needs. New accounts receive $0.10 of credit as a gift on sign-up, which is enough to send a small test request and watch the pipeline run end to end.

In your client, set the base URL to https://api.vanaapi.com and paste the key you just created. If your client speaks the OpenAI format, keep that setting as it is: GPT-6.1 Sol answers on /v1/chat/completions and on /v1/responses. The model id is gpt-6.1-sol. Codex uses the /v1/responses route.

Send one short prompt, then open the usage log in the dashboard. Every call is listed there, with the tokens that were actually charged. That first entry is the most useful part of the setup for a student, because it shows that the address, the key, the model id, and the billing all line up. Vana API is prepaid, charged per real token, and carries no monthly fee, so you can start with a small balance and let your own usage set the pace.

§ What a student run looks like

For example, say you use GPT-6.1 Sol every day for lecture notes, code review, and rewriting paragraphs. At $0.1389 per million input tokens and $0.8333 per million output tokens on Vana API, a working session is a line item rather than a subscription. Say one day sends 2 million input tokens and 1 million output tokens: 2 x $0.1389 = $0.2778 for input, and 1 x $0.8333 = $0.8333 for output. That is an example, not a quote, and it shows how the rate follows what you send.

Because the balance is prepaid and every call appears in the usage log, you can read the pattern of your own work. A week of heavy experimentation looks different from a week of light drafting, and the log shows the difference in plain numbers. For a student on a limited budget, that visibility is worth more than a long feature list: you can decide when to use a smaller model for routine tasks and when to keep a stronger one for the harder parts.

The same address covers other families when a project changes shape. GPT-6 Astra is listed at $0.6944 / $3.54 per million tokens, and Qwen3.7 Plus at $0.1278 / $0.4167. Grok 4.7, DeepSeek V4 Flash, DeepSeek V4 Pro, GLM-5.3, Kimi K3, Qwen3.8 Max, MiniMax M3, and MiMo V2.6 Pro use the OpenAI format on the same endpoint. Moving between them is a model id change, not a new integration.

§ Pricing that matches how students spend

Vana API pricing is per token, with no monthly fee on the prepaid balance. GPT-6.1 Sol costs $0.1389 / $0.8333 per million tokens (input / output) on Vana API, against an official list price of $2 / $10. GPT-6 Astra is $0.6944 / $3.54 against $10 / $50, and GPT-5.6 Sol is $0.3472 / $1.81 against $5 / $30.

If you would rather commit to a larger allowance, the GPT monthly plan is $4,000 / 30 days: $103.61 for 30 days, at most $1,200 a week, GPT models only. That fits a student whose coursework has a predictable rhythm and who wants a ceiling on weekly spend. Students who study in bursts can stay on pay-as-you-go and pay only for the bursts.

Refunds stay simple. Unused balance can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. That window is short, so a top-up is better treated as a decision for the term than a standing habit. The referral program adds another angle: 14% of what people you invite pay, 5% one level further, with earnings that can be withdrawn from $13.89.

§ Keys, balance, and a usage log you can check

Everything you need to manage the account sits in the dashboard on https://vanaapi.com: keys, balance, and the usage log. The API address at https://api.vanaapi.com only answers requests, which keeps the place where you run code separate from the place where you check what it cost. Sharing a lab machine is easier when a key can be rotated from the dashboard without editing code.

The usage log is also the simplest way to keep a study budget honest. Billing is per real token, so the log records what you actually sent rather than a rounded estimate. When a week gets expensive, the log usually points at the habit behind it: a long context pasted again and again, or a draft regenerated far more than necessary.

Five minutes is a fair estimate for the setup: create a key in the dashboard, point one client at one address, send one request, and read one log entry. After that, Vana API sits quietly underneath whatever you are building, and the only thing left to manage is how much you want to spend on it.