Claude Fable 5.1 pricing on Vana API: $3.33 per million

Money mattersEN2026-11-03

§ One address, three protocols, minimal setup

Vana API places the models it serves behind a single base address, https://api.vanaapi.com. That address answers API requests and nothing more. Keys, balance and the usage log sit on the main site, https://vanaapi.com, where you create a key and fund the account. The division is straightforward and keeps the layout predictable: one endpoint to point at, one dashboard to manage, and a short configuration path between the two.

Behind that one address, Vana API speaks three protocols.

The Anthropic Messages protocol, at /v1/messages, for Claude models.
The Gemini native protocol, at /v1beta/models/..., for Gemini models.
The OpenAI format, at /v1/chat/completions, for GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models. GPT models also answer on /v1/responses, which is what Codex uses.

That is the whole setup idea behind Vana API. The base address stays fixed, and the path plus the application you already use determine the route. There is no monthly fee on the pay-as-you-go route: you add a prepaid balance and Vana API charges per real token, so the amount you fund is the amount the tokens cost.

The benefit of this layout is that adding a model does not mean adding a second service. The same address serves all three protocol styles, and each model id is paired with the protocol it actually speaks. Minimal setup, in practice, means remembering one URL and choosing the right route for the model in front of you.

§ Claude Fable 5.1 and the application that can reach it

Claude Fable 5.1 carries the model id claude-fable-5-1, and Vana API serves it over the Anthropic Messages protocol at /v1/messages. It is reachable only from official Claude Code or Claude Desktop, version 2.1.285 or later, because those applications speak the Anthropic Messages protocol themselves. If you already work in one of them, the integration is a configuration step rather than a project.

The other two protocol groups follow their own rules, and Vana API keeps them separate instead of blending them. Gemini models answer over the Gemini native protocol from Gemini CLI, and they are not OpenAI-compatible. The OpenAI-format group, covering GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models, uses /v1/chat/completions, with GPT models additionally available on /v1/responses for Codex. One address, three protocols, and each model paired with the client that speaks its protocol.

For Claude Fable 5.1, the practical list is short. Create a key in the dashboard on https://vanaapi.com, add funds to the prepaid balance, set the base address to https://api.vanaapi.com inside Claude Code or Claude Desktop version 2.1.285 or later, and select claude-fable-5-1. That is the minimal setup, and it is why the single address is worth remembering once and reusing for every model you add afterwards.

§ Worked example: Claude Fable 5.1 next to the official list price

The following is an example, not a quotation. Say you run a workload of one million input tokens and one million output tokens for Claude Fable 5.1 over 30 days, and you want to compare the Vana API rate with the official list price of $10 / $50 per million tokens.

On Vana API, the per-million prices are $3.33 for input and $16.74 for output, so the example workload is priced as:

Input: one million tokens at $3.33 per million = $3.33.
Output: one million tokens at $16.74 per million = $16.74.
Combined rate for the example: $3.33 + $16.74.

At the official list price of $10 / $50 per million tokens, the same example is priced as:

Input: one million tokens at $10 per million = $10.
Output: one million tokens at $50 per million = $50.
Combined rate for the same example: $10 + $50.

Read the two sums side by side and the difference is visible directly in the input and output lines: ($10 − $3.33) on input and ($50 − $16.74) on output. Every figure in the example comes from the published rates for Claude Fable 5.1, so the comparison holds for any volume you scale the same way. Because Vana API bills per real token from a prepaid balance, the tokens in your workload are the tokens you pay for, and the usage log in the dashboard lists every call so the sum can be checked line by line.

This is also why the per-million framing is useful. You do not need to guess what a month of traffic will cost. Take the input volume you expect, take the output volume you expect, and apply $3.33 and $16.74. The official list price applies $10 and $50 to the same two volumes. The gap between those two pairs of rates is the value Vana API returns on Claude Fable 5.1, expressed in the same units on both sides.

§ A second example: a repeated daily pattern

Here is another example, this time with a repeated daily pattern instead of a single monthly workload. Say you run this every day: one million input tokens and one million output tokens for Claude Fable 5.1, every day for 30 days.

On Vana API, each day of that example is priced as:

Input: one million tokens at $3.33 per million = $3.33.
Output: one million tokens at $16.74 per million = $16.74.
Daily rate for the example: $3.33 + $16.74, repeated for 30 days.

The same daily pattern at the official list price of $10 / $50 per million tokens:

Input: one million tokens at $10 per million = $10.
Output: one million tokens at $50 per million = $50.
Daily rate for the same example: $10 + $50, repeated for 30 days.

In this second example, the comparison again reduces to the difference between the two input rates and the two output rates: ($10 − $3.33) and ($50 − $16.74), applied to each day of the pattern. The arithmetic stays simple because Vana API prices input and output separately and charges per real token, so a daily pattern can be read as one daily sum rather than an estimate built from averages.

The practical use of the second example is planning. With a prepaid balance, you know the per-million rates before the first call, and the usage log on https://vanaapi.com shows what each call actually consumed. There is no monthly fee on this route, so a quiet week and a busy week both settle at the same per-token rates.

§ Balance, plans and what the dashboard shows

Vana API uses a prepaid balance charged per real token, with no monthly fee on that route. Keys are created in the dashboard, and every call is listed in the usage log, so the examples above can be verified rather than estimated. A gift of $0.10 of credit is added on sign-up, which is enough to make a first small request and confirm that the address, the key and the model id are all in the right order.

There are also monthly plans for Claude models only. The plan covering up to $1,000 / 30 days costs $64.72 for 30 days and allows at most $320 a week. The plan covering up to $2,400 / 30 days costs $153.61 for 30 days and allows at most $750 a week. Both are scoped to Claude models only, so a workflow built on Claude Fable 5.1 can run either on prepaid pay-as-you-go billing or on one of these monthly allowances, depending on how steady the usage is.

Refunds apply to unused balance: it can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. The referral programme adds a second line to the economics, with 14% of what people you invite pay and 5% one level further, and earnings that can be withdrawn from $13.89.

§ Setup in short steps

Putting Claude Fable 5.1 on Vana API comes down to a short sequence.

Open https://vanaapi.com and create a key in the dashboard.
Add funds to the prepaid balance, or choose a Claude monthly plan.
In Claude Code or Claude Desktop, version 2.1.285 or later, set the base address to https://api.vanaapi.com and select the model id claude-fable-5-1.
Send a request, then check the usage log on https://vanaapi.com to confirm the tokens and the charge.

That is the shape of Vana API: one address, three protocols, minimal setup. The Claude Fable 5.1 example shows $3.33 + $16.74 per million input and output against $10 + $50 at the official list price, and the second example repeats the same two sums for a daily pattern over 30 days. Vana API keeps the configuration short and the arithmetic in published per-token rates, so the value is visible in the numbers rather than in the wording.