Help Center

Help Center product

Eazo CreatorEazo App

Documentation language

ENEnglish中中文
Browse documentation

Configure AI Models for an App

Configure AI models, Eazo billing, your own API keys, and usage for an App.

On this page

App AI means AI used by visitors inside the App you create. It is separate from talking with the Agent in Eazo Creator.

Eazo includes official model access and usage billing for App visitors. When an App needs AI conversation, generation, analysis, recommendations, or image understanding, the Agent identifies the capability and displays a Configure AI model card after design confirmation and before the full build. With an Eazo official model, you do not need a provider account, API Key, or server integration. Select a model and understand its Credits usage. In an existing App, you can also ask the Agent to change models, add AI capabilities, use your own Key, or disable AI.

The configuration card offers three choices:

  • Do not use an AI model: continue with ordinary App functionality;
  • Use Eazo AI service: visitors need no API Key, and usage is deducted from the creator's Credits;
  • Use my own API key: the creator pays the model provider directly.

Configure AI model card

Choose Do not use an AI model, Use Eazo AI service, or Use my own API key.

Available models and capabilities such as text, image understanding, and image generation follow the current configuration card. You can also review Models and built-in AI capabilities supported by Eazo.

Billing for Eazo official models

Charges occur only when the App actually calls a model. Saving a configuration alone does not create usage charges. Text and vision are billed by input and output Tokens. Image generation is billed by actual image tokens, and the rates page shows an approximate cost per image. Video, speech-to-text, and text-to-speech are billed by duration; the rates page shows approximate per-second or per-minute costs, and the actual charge depends on length and resolution. Follow Settings > App usage > Model pricing rates for the current rates and billing method.

ProviderModelVisionInput (Credits/1M Tokens)Output (Credits/1M Tokens)Context length
xAIGrok 4.3✅41.6783.331M
GoogleGemma 4 E2B✅1.332.67128K
GoogleGemma 4 31B✅4.6713.33256K
GoogleGemma 4 26B-A4B✅4.3313.33256K
GoogleGemma 3 4B IT✅1.332.67128K
GoogleGemma 3 12B IT✅3.009.67128K
GoogleGemma 3 27B IT✅7.6712.67128K
OpenAIGPT-5.4✅91.67550.00272K
OpenAIGPT OSS 20B❌2.3310.00128K
OpenAIGPT OSS 120B❌5.0020.00128K
OpenAIGPT OSS Safeguard 20B❌2.336.67128K
OpenAIGPT OSS Safeguard 120B❌5.0020.00128K
AnthropicClaude Haiku 4.5✅36.67183.33200K
DeepSeekDeepSeek V3.2❌20.6761.67164K
DeepSeekDeepSeek-V3.1❌19.3356.00128K
QwenQwen3 Coder Next❌16.6740.00256K
QwenQwen3-Coder-30B-A3B-Instruct❌5.0020.00256K
QwenQwen3 Coder 480B A35B Instruct❌15.0060.00128K
QwenQwen3 235B A22B 2507❌7.3329.33256K
QwenQwen3 VL 235B A22B✅17.6788.67256K
QwenQwen3 Next 80B A3B❌4.6740.00256K
QwenQwen3 32B❌5.0020.0032K
Moonshot AIKimi K2.5✅20.00100.00256K
Moonshot AIKimi K2 Thinking❌20.0083.33256K
MiniMaxMiniMax M2.5❌10.0040.00196K
MiniMaxMiniMax M2.1❌10.0040.00196K
MiniMaxMiniMax M2❌10.0040.001M
Z.AIGLM 5❌33.33106.67200K
Z.AIGLM 4.7❌20.0073.33203K
Z.AIGLM 4.7 Flash❌2.3313.33203K
Mistral AIDevstral 2 123B❌13.3366.67256K
Mistral AIMagistral Small 2509✅16.6750.00128K
Mistral AIMistral Large 3✅16.6750.00256K
Mistral AIMinistral 14B 3.0✅6.676.67128K
Mistral AIMinistral 3 8B✅5.005.00128K
Mistral AIMinistral 3B✅3.333.33128K
Mistral AIVoxtral Small 24B 2507❌3.3310.0032K
Mistral AIVoxtral Mini 3B 2507❌1.331.3332K
NVIDIANemotron 3 Super 120B❌5.0021.67256K
NVIDIANemotron Nano 3 30B❌2.008.00256K
NVIDIANemotron Nano 12B v2 VL BF16✅6.6720.00128K
NVIDIANemotron Nano 9B v2❌2.007.67128K
WriterPalmyra Vision 7B✅5.0020.004K

Model availability and rates can change. Follow the information shown in the App AI configuration interface.

When Credits run out, Apps that are enabled and use an Eazo official model pause AI usage. After a top-up, Apps that remained enabled resume automatically. An App you disabled manually does not turn itself back on.

Use my own API key

In the card's second step, choose a provider first, then choose or enter a model name for that provider, and finally enter the API Key. For a known provider, the endpoint is filled automatically and stays hidden; switching providers clears the Key when the endpoint or API format changes. If your provider is not listed, choose Custom and enter the HTTPS Base URL yourself (do not append paths like /chat/completions). The API format is inferred from the provider preset or custom endpoint markers. You can configure AI before creating the App, without starting a preview first. Confirmation applies the configuration to the current project and preserves it when the App is created. Reopening the card requires entering the configuration again. Applying the configuration and verifying the model connection are separate results. The Key is not sent as a normal Chat message or placed in browser code.

API Key configuration card

Choose a provider and model name, then enter the Key. Known providers hide the endpoint; Custom shows a Base URL field. The screenshot is pending an update.

Do not send a complete Key in a normal Chat message. Calls made with your Key do not consume Eazo Credits for App AI; you manage the provider's charges and quotas.

Review and manage App AI usage

Open More > Settings > App usage to review:

  • The current model for each App;
  • Total, last-30-day, and last-7-day Credits usage;
  • Call count;
  • Credits used, plus input and output Tokens for text models;
  • Current model rates, including billing for image, video, and speech.

App AI usage

App usage in Settings shows models, consumption, calls, Tokens, and rates for each App.

You can disable Eazo AI for one App. New calls stop immediately while history remains. Re-enable it to let visitors use AI again.

Change an App's AI model

Return to the Project Chat and describe the goal. A model-only change usually updates configuration, while adding a new AI capability may require App changes. Test again and update the published version afterward.

Related guides