Capy Compute
a product of Capy Ventures LLC.
Buy inference with SOL
Top up from a wallet, take an API key, point any OpenAI client at it. No card, no account, no monthly minimum. You are charged for the tokens you actually use, plus 10%.
Connect a Solana wallet to buy credit and take an API key.
Models and prices
Per million tokens, before the 10% margin. These are the upstream prices, shown so the margin can be checked rather than taken on trust.
| Model | Input | Output |
|---|---|---|
Claude Sonnet 4.5 Best general-purpose reasoning in the catalogue. The default for anything hard. anthropic/claude-sonnet-4.5 | $3.00 | $15.00 |
Claude Haiku 4.5 Fast and cheap, and still good at instructions. The right choice for volume. anthropic/claude-haiku-4.5 | $1.00 | $5.00 |
GPT-4.1 mini Inexpensive, wide knowledge, long context. Good for extraction and classification. openai/gpt-4.1-mini | $0.40 | $1.60 |
Gemini 2.5 Flash The cheapest way to read a very long document end to end. google/gemini-2.5-flash | $0.30 | $2.50 |
Llama 3.3 70B Open weights at the lowest price here. Fine for drafting and summarising. meta-llama/llama-3.3-70b-instruct | $0.12 | $0.30 |
Using it
The endpoint is OpenAI-shaped, so an existing client needs two strings changed and nothing else.
curl https://capycompute.com/v1/chat/completions -H "Authorization: Bearer capy_..." -H "Content-Type: application/json" -d '{
"model": "anthropic/claude-haiku-4.5",
"messages": [{ "role": "user", "content": "Say hello" }]
}'from openai import OpenAI
client = OpenAI(
base_url="https://capycompute.com/v1",
api_key="capy_...",
)Every response carries a capy block with what the request cost and what is left, so a client can watch its own balance without a second call. When the balance runs out the endpoint returns 402.