Claude Opus 5.5 pricing: one address, three protocols with Vana API

§ One address, three protocols

Vana API gives you one API address, api.vanaapi.com, and three ways to reach it: Anthropic Messages (/v1/messages), the Gemini native protocol (/v1beta/models/...) and the OpenAI format (/v1/chat/completions). GPT models also answer on /v1/responses. That is the whole idea behind the setup. You point a client at one address, choose the protocol that client already speaks, and start working.

The dashboard lives somewhere else from the API address, and that separation keeps things easy to reason about. Keys, balance and the usage log are managed at vanaapi.com, while api.vanaapi.com answers API requests only. The main site is where you look; the API address is what you configure.

For Claude Opus 5.5 the answer is short. It runs on the Anthropic Messages protocol, and it is used from official Claude Code or Claude Desktop, version 2.1.285 or later, because those applications speak /v1/messages themselves. There is no translation layer to install, no format to rewrite and no second endpoint to memorise before the model answers.

Three protocols on one address also means you do not juggle separate accounts per model family. When your work shifts from one model to another, the address stays the same and the protocol changes with it. That is the shape of Vana API in a single sentence: one address, three protocols, minimal setup. The rest of this checklist tells you which protocol to pick for the model you want, how to configure it, and how the money works.

§ Checklist: point Claude Opus 5.5 at the right protocol

First, create a key in the dashboard at vanaapi.com. Keys are made there, and the key you create is the value you paste into your client configuration. Nothing is generated on the API side.

Second, use the API address api.vanaapi.com. This is the only endpoint you need for the protocol you have chosen, and it is where the request goes for every call.

Third, run official Claude Code or Claude Desktop at version 2.1.285 or later. Those two applications speak Anthropic Messages (/v1/messages) themselves, so the request format they produce already matches what Vana API expects. If the application you are using is older than 2.1.285, bring it up to that version or later before you start sending requests.

Fourth, set the model id to claude-opus-5-5. That is the identifier Vana API uses for Claude Opus 5.5, and it goes in the model field of the request. The model name and the protocol are the two values that decide where a request lands.

Fifth, check the usage log in the dashboard after your first call. Every call is listed there, with the spend attached to it, so you can see what a request cost and compare it with your own estimate.

That is the complete path for Claude Opus 5.5: dashboard for the key, one address for the traffic, one model id, and a log that shows the spend. There is no separate account to open for this model and no second service to sign up with.

§ When you want a different model family

The other two protocols exist so you can keep the same address when your model choice changes. The decision is a protocol decision, and each family has a clear home.

Gemini models are reached over the Gemini native protocol (/v1beta/models/...), and they are used from Gemini CLI. The Gemini native protocol is not OpenAI-compatible, so the request shape stays native and the client stays Gemini CLI. That is the configuration for this family, and it keeps the endpoint count at one.

For GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo, the OpenAI format (/v1/chat/completions) is the path. GPT models also answer on /v1/responses, which is the route Codex uses. Whatever tool you already point at an OpenAI-compatible endpoint can be pointed at this one instead, because the format and the address are already the ones it expects.

So the choice comes down to what your client speaks natively. Anthropic Messages for Claude, the Gemini native protocol for Gemini, the OpenAI format for the rest, with /v1/responses available for GPT workflows. One address covers all three, and in most cases the client you already use tells you which protocol it expects without any documentation search.

§ What Claude Opus 5.5 costs

Claude Opus 5.5 is priced at $1.46 per million input tokens and $6.94 per million output tokens on Vana API. The official list price for the same model is $4 and $20, so both sides of the token count come out lower on Vana API.

For a sense of where Opus 5.5 sits in the Claude range, Claude Fable 5.1 is $3.33 and $16.74 on Vana API against $10 and $50 officially, Claude Haiku 4.5 is $0.3472 and $1.60 against $1 and $5, and Claude Sonnet 5.5 is $1.18 and $5.76 against $2 and $10. Opus 5.5 sits in the middle of that spread, which makes it a reasonable default when you want a capable model without paying at the top of the range.

Here is an example, say you run this every day: one call that uses 100,000 input tokens and 20,000 output tokens. At $1.46 per million input tokens, the input part works out to $0.146. At $6.94 per million output tokens, the output part works out to $0.1388. Together that is $0.2848 for the call, and the same figure appears in the dashboard usage log as a listed call.

Per-token pricing also means you are not paying for capacity you do not use. There is no monthly fee on the pay-as-you-go balance, so a quiet week costs what your requests cost and nothing beyond that.

If your usage is steadier and larger, the monthly plans put a ceiling on the same per-token model. The Claude plan up to $1,000 / 30 days costs $64.72 for 30 days, with at most $320 a week, and it covers Claude models only. The Claude plan up to $2,400 / 30 days costs $153.61 for 30 days, with at most $750 a week, also Claude models only. For GPT there is a plan up to $4,000 / 30 days at $103.61 for 30 days, with at most $1,200 a week, for GPT models only.

§ Balance, plans, referral and refunds

Vana API uses a prepaid balance. You top up, calls are charged per real token, and there is no monthly fee attached to the balance itself. Balance and usage sit in the dashboard next to the keys you created, so spending is visible without asking anyone for a report.

New accounts receive $0.10 of credit as a gift on sign-up. That is a small starting point, but it is enough to make a first request and see how the usage log records it.

Referrals add a second layer to the same balance. You receive 14% of what people you invite pay, and 5% one level further, with earnings that can be withdrawn from $13.89. The percentages are calculated on what your invitees actually pay, so the figure follows real usage rather than a fixed tier.

Refunds are straightforward as well. Unused balance can be refunded within 3 days of the top-up, minus what was used and the payment processor's fee. Keeping a close eye on the usage log in the days after a top-up makes that window easy to manage.

Put together, the value picture is plain. One address, three protocols, minimal setup, per-token billing with no monthly fee, and a dashboard that shows every call. For Claude Opus 5.5, the checklist stays short: make a key at vanaapi.com, point official Claude Code or Claude Desktop 2.1.285 or later at api.vanaapi.com, choose claude-opus-5-5, and read the log afterwards.