Recursive inference for companies to train their own agents and models.

Run your product on our inference cloud. Every call is captured as training data you own, so the model you train is yours to serve.

Up to $25 of credit on signup. No card.

For developers

One API.Every model.

Point the OpenAI SDK at Allocate and call any model by name. Every call lands in your own database, and the served model comes back in a header.

175Models
30 msOverhead budget
3Regions
main.py
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.allocate.network/v1",
    api_key=os.environ["ALLOCATE_API_KEY"],
)
reply = client.chat.completions.create(
    model="moonshotai/kimi-k3",
    messages=[{"role": "user", "content": "Summarise this ticket."}],
)
print(reply.choices[0].message.content)
Models

Open weight and frontier.One catalog.

69 open weight language models you can train into your own, next to the frontier labs. One key, one bill.

Language91Image31Video37Speech6Transcription7Embedding2Rerank1
models in the catalog
you can train into your own
regions for compute and data

Latest

All posts

Choose how to get started

Build on your own

Sign up, create a key, and send your first request in a minute. Capture starts with that request.

  • One key for every model in the catalog
  • 5 agent computer hours every month
  • Every call captured in RecursiveDB
  • Metered per second, no seat licence

Enterprise

Your own cell, in your own region, and a model your customers use that you own outright.

  • A single tenant cell in your region
  • Weights and data you can export
  • Security review and a signed DPA
  • A direct line, not a ticket queue