Cursor pricing on Vana API: $0.3472 per million input tokens

§ One address, three protocols, one Cursor workflow

When an agency evaluates an API relay for Cursor, the practical question is how much setup sits between the decision and the first working request. With Vana API, the answer is short: one address, three protocols and a small amount of configuration. Every request goes to https://api.vanaapi.com. That address answers API requests only, while keys, balance and the usage log are handled in the dashboard on https://vanaapi.com. Keeping those two roles separate means your team can manage spend and access in one place without changing the endpoint your editors already use.

Three protocols are served from that single address. The OpenAI format at /v1/chat/completions is the one that matters for Cursor, and it covers GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models. GPT models also answer on /v1/responses, the endpoint Codex uses. The other two protocols on the same address are the Anthropic Messages protocol and the Gemini native protocol, and each remains with its own client. For a Cursor-focused agency, the point is simpler: you are not assembling a collection of hosts and billing relationships, because Vana API already puts the protocols you need behind one endpoint.

That structure also makes the pricing conversation straightforward. Vana API uses a prepaid balance charged per real token, with no monthly fee. There is no seat count to negotiate and no contract to interpret before a single request is sent. The numbers below are published per million tokens, so an agency can compare a Cursor workflow against an official list price using the same units on both sides.

§ Setting up Cursor against Vana API

The setup is deliberately minimal. First, create an account on the Vana API site and open the dashboard. Keys are created in the dashboard, so the key you copy into Cursor originates from the same place that shows your balance and your usage log. Second, add credit to that balance. New accounts receive $0.10 of credit as a gift on sign-up, which is enough to confirm that the connection works before any larger top-up.

Third, point Cursor's OpenAI-compatible provider at https://api.vanaapi.com and paste the key you created. Fourth, choose a model from the list that Cursor can use and enter its model id. The Cursor-compatible set includes GPT-5.6 Sol (gpt-5.6-sol), GPT-6.1 Sol (gpt-6.1-sol), Grok 4.7 (grok-4.7), DeepSeek V4 Flash (deepseek-v4-flash), DeepSeek V4 Pro (deepseek-v4-pro), GLM-5.3 (glm-5.3), Kimi K3 (kimi-k3), Qwen3.7 Plus (qwen3.7-plus), Qwen3.8 Max (qwen3.8-max), MiniMax M3 (MiniMax-M3) and MiMo V2.6 Pro (MiMo-V2.6-Pro). Because these models all travel over the OpenAI format at /v1/chat/completions, a single provider entry in Cursor covers the whole range.

That is the whole configuration path. There is no separate gateway to install, no per-protocol account to create and no second balance to fund. Once the base URL and key are in place, switching between a GPT model and a DeepSeek, Qwen or MiniMax model is a change of model id rather than a change of infrastructure. For an agency running several client accounts, that consistency is often more valuable than any individual price point.

§ How an example workflow runs from request to invoice

The following is an example workflow, and the figures are illustrative arithmetic based on the published rates. Say you run this every day: a small agency uses Cursor for code review across three client repositories, and the technical lead wants a predictable per-project cost. You decide to start with GPT-5.6 Sol for the review pass and keep the rest of the workflow on the same provider entry in Cursor.

Say one project produces 1,000,000 input tokens and 1,000,000 output tokens in a month. On Vana API, GPT-5.6 Sol is priced at $0.3472 / $1.81 per million tokens, so that example month costs $0.3472 plus $1.81, which is $2.1572 in total. The official list price for the same model is $4 / $20, so the same volume priced at list would be $24. That comparison is the whole point: the same token counts, the same model id and the same Cursor setup, with the difference coming from the rate Vana API charges.

Now say a lighter project runs on GPT-6.1 Sol instead. At $0.1389 / $0.8333 per million tokens, 10,000,000 input tokens alone would come to $1.389, calculated as $0.1389 multiplied by 10. A client that needs more reasoning capacity on a smaller volume can move up to a different model without changing endpoints, because Vana API already speaks the OpenAI format that Cursor expects.

The reporting step is just as short. Every call is listed in the usage log in the dashboard, so at the end of the month you can reconcile what was spent against what was delivered rather than estimating from memory. Because billing is prepaid and charged per real token, the balance itself is the control. If a project starts consuming more than expected, you can see it in the log while it is happening. The agency in this example can then prepare its client invoice from recorded usage, and the technical lead can see which model produced it.

§ What the Cursor-compatible models cost on Vana API

Pricing for the models Cursor can use is published per million tokens, with input and output shown separately. The full list is short enough to keep on an internal page:

• GPT-5.6 Sol (gpt-5.6-sol): Vana API $0.3472 / $1.81, official list price $4 / $20
• GPT-6 Astra (gpt-6-astra): Vana API $0.6944 / $3.54, official list price $10 / $50
• GPT-6.1 Sol (gpt-6.1-sol): Vana API $0.1389 / $0.8333, official list price $2 / $10
• Grok 4.7 (grok-4.7): Vana API $0.4861 / $1.53
• DeepSeek V4 Flash (deepseek-v4-flash): Vana API $0.2778 / $1.04
• DeepSeek V4 Pro (deepseek-v4-pro): Vana API $1.11 / $3.33
• GLM-5.3 (glm-5.3): Vana API $0.5556 / $1.67
• Kimi K3 (kimi-k3): Vana API $1.04 / $5.21
• Qwen3.7 Plus (qwen3.7-plus): Vana API $0.1278 / $0.4167
• Qwen3.8 Max (qwen3.8-max): Vana API $0.6944 / $2.08
• MiniMax M3 (MiniMax-M3): Vana API $0.1389 / $0.5556
• MiMo V2.6 Pro (MiMo-V2.6-Pro): Vana API $0.2083 / $0.3472

For teams that prefer a fixed arrangement, Vana API also offers one monthly plan: GPT up to $4,000 / 30 days for $103.61 for 30 days, with at most $1,200 a week and GPT models only. Everything else runs on the prepaid balance with no monthly fee, which suits agencies whose volume changes from month to month and who would rather not commit to a subscription for models they use only occasionally.

§ Balance, refunds and a straightforward dashboard

The financial mechanics of Vana API stay visible. You top up, the balance is charged per real token, and the usage log records every call. If a top-up turns out to be larger than a project needs, 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 worth noting in an agency's procurement process, since it determines when a balance decision should be made rather than after the fact.

There is also a referral programme for agencies that bring in other teams. You receive 14% of what people you invite pay, and 5% one level further. Earnings can be withdrawn from $13.89. For a studio that already recommends its stack to other studios, that turns the existing network into a measurable line rather than an informal favour.

Put together, the pitch for an agency working with Cursor is compact: one address at https://api.vanaapi.com, three protocols on it, and a setup that consists of creating a key, adding credit and entering a model id. Vana API publishes its rates per million tokens, lists every call in the dashboard and charges nothing monthly on the pay-as-you-go balance. For a team that wants a low-friction API relay for Cursor, the remaining work is choosing the model that fits the task and the budget.