LobeHub Credit Billing Rules
This feature is available only in LobeHub Cloud
LobeHub Cloud uses credits to measure usage of its built-in model service. Prices vary by model and by input or output type, so the same number of tokens can consume different amounts of credits.
Credit Conversion
For calculation purposes, LobeHub converts USD 1 of model usage into 1,000,000 credits.
For example, if a model costs $3 / 1M input tokens, each input token corresponds to 3 credits. Using 1,000 input tokens consumes approximately 3,000 credits.
The final deduction is based on the completed request's usage and the model price in effect at that time. You can view current prices in the model details.
Text Conversation Billing
Text models commonly charge separately for:
- Input tokens: Information sent to the model, such as the current message, system prompt, conversation history, file content, and tool results.
- Output tokens: The response generated by the model, including reasoning content for some models.
- Cache reads and writes: Models that support prompt caching apply their own cache rates, which usually differ from standard input rates.
Text Model Pricing Examples
The following examples show pricing for selected text models. All credit amounts are per 1M tokens. A range indicates tiered pricing; the applicable tier depends on conditions such as input length.
| Model | Context window | Input | Output | Cache read |
|---|---|---|---|---|
| GPT-6 Astra | 1.05M | 10M–20M credits | 50M–75M credits | 1M–2M credits |
| GPT-5.6 Sol (list price) | 1.05M | 5M–10M credits | 30M–45M credits | 0.5M–1M credits |
| GPT-5.6 Terra | 1.05M | 2M–4M credits | 12M–18M credits | 0.2M–0.4M credits |
| GPT-5.6 Luna | 1.05M | 0.2M–0.4M credits | 1.2M–1.8M credits | 0.02M–0.04M credits |
| Claude Fable 5.1 | 1M | 10M credits | 50M credits | 0.25M credits |
| Claude Fable 5 | 1M | 10M credits | 50M credits | 1M credits |
| Gemini 3.8 Flash | 1M | 0.75M credits | 3.75M credits | 0.075M credits |
| DeepSeek V4 Pro | 1.05M | ~0.632M credits | ~1.896M credits | ~0.021M credits |
| Grok 4.6 | 500K | 2M–4M credits | 6M–12M credits | 0.5M–1M credits |
| GLM-5.3 | 1.05M | ~1.124M credits | ~3.933M credits | ~0.281M credits |
| Kimi K3 | 1M | 3M credits | 15M credits | 0.3M credits |
GPT-5.6 Sol may currently display promotional rates. For durable reference, the table uses the list prices shown with strikethrough in the supplied screenshot: 5M–10M input, 30M–45M output, and 0.5M–1M cache-read credits per 1M tokens. Your actual charge follows the price shown in the model details when you use it.
GPT-6 Astra and the GPT-5.6 family change tiers at 272K input tokens. Grok 4.6 changes tiers at approximately 200K input tokens. DeepSeek and GLM list prices in CNY; the table converts them to credits using the codebase rate of USD 1 = CNY 7.12 and rounds the results.
For example, a Gemini 3.8 Flash request with 10,000 standard input tokens and 2,000 output tokens costs:
- Input:
10,000 × 0.75 = 7,500credits - Output:
2,000 × 3.75 = 7,500credits - Total: approximately
15,000credits
If 8,000 input tokens are served from cache, that portion uses the lower cache-read rate and the request costs less. The final deduction is based on the actual usage reported for the completed request.
These examples are based on the current model configuration in GitHub. Model availability and prices can change when providers update their services. Check the model details for the price in effect when you use it.
As a conversation grows, each request usually includes more history and consumes more credits. Starting a new conversation when the topic changes can reduce unnecessary input.
When you use web scraping, files, knowledge bases, or tools, the related content may be sent to the model as input and counted toward the request.
Tiered and Parameter-based Pricing
Some models use tiered pricing. When the input exceeds a specified threshold, the excess portion or the entire request may use a higher rate. For example, some Gemini Pro models move to a higher pricing tier when input exceeds 200K tokens.
Some cache, image, and video models also select prices based on parameters such as cache duration, image size, resolution, or length. Check the model details for the applicable units and rates.
Image, Audio, and Video Billing
Multimodal models deduct credits using the units supported by each model. Common units include:
- Per 1M tokens
- Per 1M characters
- Per image
- Per video
- Per megapixel
- Per second
Credit usage can vary when you use different models, dimensions, quality settings, or durations for the same task.
View Prices and Usage
- Open a model's details in the model selector to view current input, output, cache, and multimodal prices.
- Open App Settings → Plans → Usage to view usage for the current billing cycle.
- If you need more credits, you can upgrade your plan, purchase a credit package, or configure your own model API key.
When you use a custom model API key, the provider charges your provider account under its own pricing rules. The usage is not deducted from LobeHub Cloud's built-in model credits.