Model Credit Consumption
When using Nexu Official models, each conversation deducts credits based on the model type and token usage. This page lists the credit consumption for all currently available official models, helping you choose the right model and plan your credit usage.
Credit Consumption by Model
Credits are charged separately for input tokens and output tokens, with different rates per model. The table below is sorted by estimated cost per conversation from lowest to highest.
How Many Credits Does a Conversation Cost?
Using a typical single-turn conversation as an example — approximately 2,000 input tokens (including system prompt and context) and 2,000 output tokens:
| Model | Rate | Est. Cost per Conversation | Conversations per 100 Credits |
|---|---|---|---|
| deepseek-v3.2 | 0.18x | ≈ 0.13 credits | ≈ 769 |
| minimax-m2.7 | 0.43x | ≈ 0.30 credits | ≈ 333 |
| gemini-3.1-flash-lite-preview | 0.50x | ≈ 0.35 credits | ≈ 285 |
| kimi-k2.5 | 0.60x | ≈ 0.42 credits | ≈ 238 |
| glm-5 | 0.86x | ≈ 0.60 credits | ≈ 166 |
| gemini-3-flash-preview (default) | 1.00x | ≈ 0.70 credits | ≈ 142 |
| mimo-v2-pro | 1.14x | ≈ 0.80 credits | ≈ 125 |
| glm-5-turbo | 1.49x | ≈ 1.04 credits | ≈ 96 |
| gpt-5.4-mini | 1.50x | ≈ 1.05 credits | ≈ 95 |
Rate indicates the cost ratio relative to the default model, gemini-3-flash-preview. For example, 0.50x means half the cost of the default model, while 1.50x means 1.5 times the cost.
These are estimates sorted by cost from lowest to highest. Actual consumption depends on conversation length, context size, and model output. Complex tasks or multi-turn conversations will use significantly more tokens.
How to Choose the Right Model
- Maximum savings: deepseek-v3.2 and minimax-m2.7 have the lowest cost per conversation, ideal for simple Q&A, casual chat, or when credits are limited.
- Balanced choice: gemini-3.1-flash-lite-preview, kimi-k2.5, and glm-5 strike a good balance between capability and cost, suitable for most everyday tasks.
- Quality first: gemini-3-flash-preview, mimo-v2-pro, glm-5-turbo, and gpt-5.4-mini offer stronger capabilities, better suited for coding assistance, deep analysis, and other demanding tasks.
How to Get More Credits
nexu offers multiple ways to earn credits. See Earn Points by Sharing for details.
FAQ
Q: How are credits deducted?
Each conversation is billed based on the actual input tokens and output tokens consumed. The sum of both is the total credit cost for that conversation.
Q: What's the difference between input and output tokens?
Input tokens include your message, system prompts, and conversation context. Output tokens are the content generated by the model. Output tokens typically cost more because generating content requires more computation.
Q: Why do different models cost so much differently for the same conversation?
Different models have different capabilities and computational costs. More powerful models (e.g., gpt-5.4-mini, glm-5-turbo) cost more per conversation, while lightweight models (e.g., deepseek-v3.2, minimax-m2.7) are extremely affordable for simpler tasks.
Q: Does BYOK (Bring Your Own Key) mode consume credits?
No. In BYOK mode, you use your own API key to call model providers directly. Fees are charged by the provider, and no nexu credits are consumed.