DeepSeek V4 Pro for reasoning and code, V4 Flash for high-volume work. OpenAI and Anthropic formats, one key.
deepseek-v4-pro-0813−49%deepseek-v4.1-flash−84%deepseek-v4-flash-0731−77%Struck-through = the provider's official API price. Prices in USD, live from our billing.
10M input + 2M output tokens on DeepSeek V4 Pro, no cache:
DeepSeek is already one of the lowest-cost frontier model families. We go further: our per-token price is shown next to DeepSeek's own published (off-peak) price, so you can check the saving yourself.
Flash models are ideal for agents, classification, extraction and batch jobs where you send millions of tokens a day.
Sign up with Google, email or Telegram and top up with crypto from $1. The same key works with Claude, GPT and GLM if you ever need a stronger model for one step.
curl https://api.apixoai.online/v1/chat/completions \
-H "Authorization: Bearer sk-...your-key" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek-v4-pro-0813",
"messages":[{"role":"user","content":"Hello"}]}'That's about 111K tokens on Claude Fable 5. Valid 72h, for Google and Telegram sign-ups. Works on 13 models, including GPT-6.1 Sol, GPT-6 Luna, Claude Fable 5.
Sign up with Google, email or Telegram. Your API key is ready instantly ($0.10 free to test for Google and Telegram sign-ups).
Paste our base URL and your key into your SDK or tool. The docs have the exact config for each one.
Top up with crypto from $1. Every request shows its tokens and exact cost.
We buy model capacity in bulk and resell it per token. Same models, same API, a smaller margin.
The struck-through prices come from each provider's official page. Compare them yourself:
Start with the free credit or a $1 top-up. Your topped-up balance never expires and every request shows its exact cost.
More answers in the docs, or contact us.
Yes, and Anthropic-compatible too: use /v1/chat/completions or /v1/messages with the same key.
Every model shows both prices. We compare against DeepSeek's off-peak price, the lower of their two rates, so the saving shown is conservative (up to 84%).
Yes: Cline, OpenCode, Roo Code, Continue and Cursor all accept a custom OpenAI-compatible endpoint.