• Chat
  • Pricing
  • Blog
  • Downloads
  • API
    • Platform overview
    • Models
    • Rankings
    • API docs
    • Platform console
  • Company
    • About
    • Contact
    • Changelog
    • Roadmap
    • Contributors
  • Chat
  • Pricing
  • Blog
  • Downloads
  • API
    • Platform overview
    • Models
    • Rankings
    • API docs
    • Platform console
  • Company
    • About
    • Contact
    • Changelog
    • Roadmap
    • Contributors
Ditfa Platform
  • Overview
  • Models
  • Compare
  • Rankings
  • Docs
Open console
Terms of ServicePrivacy PolicyRefund PolicyAcceptable UseContact

© 2026 Ditfa. All rights reserved.

TeamtemGitHubHuggingface
  • Chat
  • Pricing
  • Blog
  • Downloads
  • API
    • Platform overview
    • Models
    • Rankings
    • API docs
    • Platform console
  • Company
    • About
    • Contact
    • Changelog
    • Roadmap
    • Contributors
  • Chat
  • Pricing
  • Blog
  • Downloads
  • API
    • Platform overview
    • Models
    • Rankings
    • API docs
    • Platform console
  • Company
    • About
    • Contact
    • Changelog
    • Roadmap
    • Contributors
Ditfa Platform
  • Overview
  • Models
  • Compare
  • Rankings
  • Docs
Open console

One API for every model, billed by the token.

Get an API key Read the docs

Change one line. Keep your OpenAI SDK and your code. Use an sk-dp- key and any model from the catalog.

 import os from openai import OpenAI   client = OpenAI(Removed line: -    base_url="https://api.openai.com/v1",Added line: +    base_url="https://gateway.ditfa.com/v1",     api_key=os.environ["DITFA_API_KEY"],  # sk-dp-... )   completion = client.chat.completions.create(     model="openai/gpt-4o-mini",     messages=[{"role": "user", "content": "Write a haiku about the ocean."}], )   print(completion.choices[0].message.content)

Live prices, per million tokens

Read from the same catalog the API serves. Each price is the lowest across providers, in USD.

What happens to a request

A model can have several providers. The gateway tries up to 3 of them in priority order and moves providers that are failing for several users to the back of the line.

Your app

POST /v1/chat/completions

Ditfa gateway

Checks the key, holds credits and picks providers

  1. 1Provider 1Timeout, 408, 429 or 5xx
  2. 2Provider 2Answers
  3. 3Provider 3Not called

If a provider times out, returns 408, 429 or 5xx, or cannot be reached, the next one is tried automatically. Nothing has been sent to you yet, so fallback is invisible.

Your app sees
The answer, as if nothing failed
You pay
Tokens used, at the rate of the provider that answered

Billing you can read

Prepaid USD credits, charged per token. No subscription and no usage windows.

Prepaid credits
Top up $5 to $1,000 at a time. Credits never expire.
Priced per token
Input, output and cached input each have a listed price. Some models add a fixed fee or a minimum per request, shown on the model page.
Held, then settled
Before a paid request the gateway holds an upper bound. When the response ends, real usage is charged and the rest is released.
Free and sponsored models
Free models cost nothing. On sponsored models Ditfa pays during the offer. Limits apply to both and are listed on the model page.
Payment fees
The payment provider’s fee is added at checkout and shown before you pay.

Three steps to a first request

  1. 1

    Add credits

    Top up from $5 in the console. Eligible accounts can use free and sponsored models without a top-up.

    Open credits
  2. 2

    Create an API key

    Keys start with sk-dp-. Set limits now or add them later.

    Create a key
  3. 3

    Send your first request

    Use any OpenAI SDK with the Ditfa base URL.

    Read the quickstart

Sell your own model

Host a model yourself, set your price per million tokens and reach every developer on Ditfa Platform. Ditfa reviews every seller.

Become a provider Provider terms

Controls you can check

What each provider does with your data, and the limits you set yourself.

Data policy per provider
Model pages show, provider by provider, whether prompts are logged and whether data may be used for training.
Provider opt-outs
Allow only verified providers, or only providers that do not log prompts, for your whole account or for one key.
Key controls
Set per key, up to 25 active keys:
  • Spend limits by day, week, month or in total
  • IP allowlists
  • Model allowlists
  • Expiry dates

Start with one key.

Get an API key Read the docs

Your analytics choice

Allow Google Analytics to measure visits and public-page clicks so we can improve Ditfa? Optional analytics stays off until you choose. Privacy policy

Change one line. Keep your OpenAI SDK and your code. Use an sk-dp- key and any model from the catalog.

 import os from openai import OpenAI   client = OpenAI(Removed line: -    base_url="https://api.openai.com/v1",Added line: +    base_url="https://gateway.ditfa.com/v1",     api_key=os.environ["DITFA_API_KEY"],  # sk-dp-... )   completion = client.chat.completions.create(     model="aion-labs/aion-3.5",     messages=[{"role": "user", "content": "Write a haiku about the ocean."}], )   print(completion.choices[0].message.content)
Ditfa Platform models with input and output prices per 1M tokens
ModelInput / 1MOutput / 1MContext
ALAion 3.5aion-labs$3.26$6.52262K
ALAion 3.5 Miniaion-labs$0.761$1.52262K
ALAion-2.0aion-labs$0.869$1.74131K
ALAion-3.0aion-labs$3.26$6.52131K
ALAion-3.0-Miniaion-labs$0.761$1.52131K
ALAion-RP 1.0 (8B)aion-labs$0.869$1.7433K
RBeam 1.0 501B A23Breflection$0.272$1.091.05M

See all 450 models Compare models

ALOne request on Aion 3.5
2,000 input tokens
at $3.26 / 1M
500 output tokens
at $6.52 / 1M
You pay
$0.00978

A $5 top-up covers about 511 of these.

Live lowest price, with the same rounding as billing. You are billed at the rate of the provider that answers.

Models
450
Model makers
67
Most providers on one model
1
Free or sponsored now
19