← All models

llama3.3-70bself-hostable

Run llama3.3-70b through Flow AI's OpenAI-compatible API. We route to the cheapest provider hosting it and fail over automatically — you pay the live clearing price plus 2.5%.

Use it

from openai import OpenAI
client = OpenAI(base_url="https://api.flowaiapi.com/v1", api_key="fa-...")
r = client.chat.completions.create(
    model="llama3.3-70b",          # or "auto" to let Flow AI pick the cheapest capable model
    messages=[{"role": "user", "content": "..."}],
)

How Flow AI makes llama3.3-70b cheaper

Flow AI runs an open market for inference: it clears each request at the lowest price across every provider hosting llama3.3-70b — direct APIs, inference providers, flat-rate subscriptions, and self-hosted nodes — and bills you the true (prompt-cache-aware) cost plus a flat 2.5%. The published rate is always the ceiling. See the docs or all models.