Claude Code pricing on Vana API: $0.3472 per million tokens

§ One address, three protocols

Most questions about Claude Code on Vana API start with a practical detail: there is a single API address to work with. That address is https://api.vanaapi.com, and it answers API requests only. Keys, balance and the usage log live in the dashboard on the main site, https://vanaapi.com. There is no second endpoint to remember and no separate account per model family.

The same address serves three protocol styles, and the protocol decides which client you use. Claude models speak the Anthropic Messages protocol (/v1/messages). Gemini models speak the Gemini native protocol (/v1beta/models/...). GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models use the OpenAI format (/v1/chat/completions). GPT models also answer on /v1/responses, which is what Codex uses.

• https://api.vanaapi.com — the one address that answers API requests.
• https://vanaapi.com — the dashboard for keys, balance and the usage log.
• Three protocols on that one address, each matched to the client that speaks it.

Because the routing is by protocol rather than by separate hosts, setup stays small. You create a key in the dashboard, add the address to your client, and choose a model id from the list. Vana API does not ask you to run a local proxy or maintain a separate endpoint for every family of models.

§ Does Claude Code work with Vana API, and what version do I need?

Yes, with a specific version requirement. Claude models on Vana API are reached from official Claude Code or Claude Desktop, version 2.1.285 or later. Those clients speak the Anthropic Messages protocol themselves, so the request format matches what they already send. There is no translation layer to configure and no custom request builder to write.

There is a clear boundary around this. Claude models speak the Anthropic Messages protocol (/v1/messages) themselves. They do not work from any other client. If you use a different tool for the rest of your work, keep that tool on the OpenAI-format models and use Claude Code or Claude Desktop when you want a Claude model.

The model ids you can enter are the ones listed for Vana API, including claude-fable-5-1, claude-haiku-4-5-20251001, claude-opus-5-5 and claude-sonnet-5-5. In practice, the sequence is short: sign up on the main site, create a key in the dashboard, point Claude Code or Claude Desktop at https://api.vanaapi.com, and pick a model id.

One note on what Vana API is. It is an API relay service, and Claude models are available on Vana API's own address through the Anthropic Messages protocol. The dashboard lists every call in the usage log, so you can check which model answered, how many tokens were processed and what the call cost.

§ What about Gemini and the OpenAI-format models?

Gemini models on Vana API use the Gemini native protocol (/v1beta/models/...). Gemini CLI is the client that speaks it. They are not OpenAI-compatible, so the OpenAI format is not the right route for them.

The OpenAI-format group is the broadest one. GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models are available over /v1/chat/completions, and they work from Cursor, Cline, Roo Code, OpenCode, Cherry Studio, any OpenAI SDK or cURL. If you prefer a terminal tool or a script, that path is already covered. GPT models additionally answer on /v1/responses, which is the endpoint Codex uses.

For example, say you send a daily report from a script that uses an OpenAI-format model. You keep that script pointed at https://api.vanaapi.com. Separately, say you want to draft longer pieces in Claude Code. You use the same Vana API key and the same prepaid balance for both, because the address is shared and only the protocol differs.

That shared key is worth pausing on. One key in the dashboard can serve several clients, as long as each client is on a protocol it supports. You do not have to manage a key per tool, and you do not have to track a balance per tool either.

§ What does Claude Code cost on Vana API?

Vana API charges per real token, and the per-million-token rates for the Claude models sit below the published list prices shown alongside them. Input and output are priced separately.

• Claude Fable 5.1 (claude-fable-5-1): Vana API $3.33 / $16.74, official list price $10 / $50.
• Claude Haiku 4.5 (claude-haiku-4-5-20251001): Vana API $0.3472 / $1.60, official list price $1 / $5.
• Claude Opus 5.5 (claude-opus-5-5): Vana API $1.46 / $6.94, official list price $4 / $20.
• Claude Sonnet 5.5 (claude-sonnet-5-5): Vana API $1.18 / $5.76, official list price $2 / $10.

For example, say you send 1,000,000 input tokens and 1,000,000 output tokens to Haiku 4.5 in a month. At Vana API rates of $0.3472 and $1.60, that is $0.3472 plus $1.60, or $1.9472 for those two million tokens. The same volume at the listed $1 / $5 would be $1 plus $5, or $6. The gap comes from the rates themselves, and the usage log records each call, so the arithmetic can be checked against real entries rather than estimates.

Monthly plans are also available, and each one is limited to a single model family. The Claude plan of up to $1,000 / 30 days costs $64.72 for 30 days, with at most $320 a week, Claude models only. The Claude plan of up to $2,400 / 30 days costs $153.61 for 30 days, with at most $750 a week, Claude models only. The GPT plan of up to $4,000 / 30 days costs $103.61 for 30 days, with at most $1,200 a week, GPT models only.

Which one fits depends on how you work. If your usage is steady and you want a predictable ceiling, a plan gives you one. If usage varies from week to week, prepaid balance and per-token billing keep the cost tied to the tokens actually processed.

§ How billing, keys and credits work

Vana API uses a prepaid balance. There is no monthly fee on the pay-as-you-go side, and calls are charged per real token. Every call appears in the usage log in the dashboard, where you can review the model, the volume and the cost. Keys are created in the dashboard as well, so key rotation and balance checks happen in the same place.

New sign-ups receive a $0.10 credit gift. That is small by design: it is enough to make a first test call and see how the setup behaves before committing to a larger top-up.

The referral programme pays 14% of what people you invite pay, and 5% one level further. Earnings can be withdrawn from $13.89. If you would rather keep things simple, this is an optional extra rather than part of the setup.

Refunds are straightforward within the stated window. Unused balance can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. The practical takeaway is that a first top-up can stay modest while you confirm that your client, your model id and the protocol all line up.

Put together, the arrangement is compact: one address at https://api.vanaapi.com, three protocols, a key from the dashboard, and per-token pricing you can inspect line by line. Vana API keeps the moving parts to that minimum, and the numbers above are the ones you can verify in the usage log.