Models

350+ models. Two tiers.

Every model with its tier and price, in xUSD per 1M tokens with the tier fee included. Web search is available on Anonymous models only, for +0.03 xUSD per message.

Private · open models · +20%Anonymous · closed models · +15%

18 availableSearch models

Model	Tier	Context	In	Out	Ask

Sample list: 18 of 350+ models. Sample prices, live values in the app. In and out: xUSD per 1M tokens.

Which models are on xanai?

xanai gives you 350+ AI models in one chat app and one API, on BNB Chain. They sit in two tiers.

Private runs open models inside a TEE hardware enclave, for a 20% tier fee. Families include DeepSeek, Qwen, Llama, Kimi, GLM, gpt-oss, Gemma and Mistral.

Anonymous gives you closed models such as Claude, GPT, Gemini, Grok and Mistral Large, for a 15% tier fee. Your identity is hidden from the provider and only zero-data-retention endpoints are used.

Every answer is paid in xUSD and gets a signed receipt. See per-message pricing in xUSD, call any model through the OpenAI-compatible API, or read how the two tiers work.

The tiers

Where each model runs.

Per questionThe tier comes with the model you pick. Both are paid from the same balance and both leave a receipt on BNB.	PrivateOpen models · TEE20%over the provider's cost	AnonymousClosed models · proxy15%over the provider's cost
Models	DeepSeek, Qwen, Llama, Kimi, GLM, Gemma, Mistral Small	Claude, GPT, Gemini, Grok, Mistral Large
Where it runs	In an attested enclave (TEE), a sealed part of the processor	At the provider, reached through the xanai proxy
Who can read your prompt	Only the enclave that runs the model	The provider, but not who sent it
Your account at the provider	Not needed: the model runs in the enclave	Not needed: xanai's account and key are used
On your receipt	Hashes, and the attestation that the enclave ran it	Hashes, and the route through the proxy
Best when	The words themselves must stay sealed	You need a closed model

PrivateOpen models · TEE20%over the provider's cost

Models: DeepSeek, Qwen, Llama, Kimi, GLM, Gemma, Mistral Small
Where it runs: In an attested enclave (TEE), a sealed part of the processor
Who can read your prompt: Only the enclave that runs the model
Your account at the provider: Not needed: the model runs in the enclave
On your receipt: Hashes, and the attestation that the enclave ran it
Best when: The words themselves must stay sealed

AnonymousClosed models · proxy15%over the provider's cost

Models: Claude, GPT, Gemini, Grok, Mistral Large
Where it runs: At the provider, reached through the xanai proxy
Who can read your prompt: The provider, but not who sent it
Your account at the provider: Not needed: xanai's account and key are used
On your receipt: Hashes, and the route through the proxy
Best when: You need a closed model

Ask any of them.

3 free messages are enough to start, and every answer leaves its signed receipt.

Open app See pricing