AI model prices
One balance for every model. Top up from $5 with crypto and pay only for what each reply or image actually uses. No subscription, no account, and credits never expire. Prices update automatically from our provider’s live rates.
Chat models
| Model | Best for | Typical message* | Input / 1M tokens | Output / 1M tokens |
|---|---|---|---|---|
| GPT-6 Luna Standard | Everyday chat, writing and coding | ~$0.0009 | $0.3 | $1.50 |
| GPT-6 Astra Premium | Advanced reasoning and complex tasks | ~$0.09 | $30.00 | $150.00 |
| Claude Sonnet 5.5 Pro | Coding, writing and analysis | ~$0.018 | $6.00 | $30.00 |
| Claude Opus 5.5 Premium | High-end coding and reasoning | ~$0.036 | $12.00 | $60.00 |
| Gemini 4 Argon Premium | Research, reasoning and complex tasks | ~$0.018 | $6.00 | $30.00 |
| Gemini 3.8 Flash Budget | Fast everyday AI assistance | ~$0.0067 | $2.25 | $11.25 |
| Grok 4.7 Pro | Chat, analysis and alternative perspectives | ~$0.012 | $6.00 | $18.00 |
| DeepSeek V4.1 Flash Budget | Affordable coding and reasoning | ~$0.0008 | $0.14 | $1.50 |
*A typical message is about 500 tokens in and 500 tokens out (roughly 350 words each way). Long chats, long answers and reasoning use more.
Image models
| Model | Best for | Per image (1K) |
|---|---|---|
| Nano Banana 2 Image | Image generation and editing | ~$0.222 |
| Seedream 5.0 Flash Image | Affordable image generation | ~$0.054 |
Credit top-ups
$5
$5 of credit
$10
$10 of credit
$25
+$1 bonus · $26 of credit
$50
+$3 bonus · $53 of credit
$100
+$8 bonus · $108 of credit
Pay with ETH or USDT on Ethereum, or SOL or USDT on Solana. You get a live quote valid for 20 minutes. Blockchain network fees are paid by you on top of the credit amount. Credits are added once the payment is confirmed on-chain.
Credits first, then chat
Every model, chat and image, uses prepaid credits. Connect a wallet, top up from $5, and start chatting. There is no subscription and nothing renews.
How billing works
Before a reply, we hold a small estimate from your balance. When the reply finishes, you are charged the actual usage reported by the model provider and the rest of the hold is returned at once. Your balance can never go below zero. If a request fails, nothing is charged.