Add Artificial Intelligence to your app
and keep it yours.
Open models behind one API. See what every call used, set the limits yourself, and know your prompts are never kept.
Start building
Put Artificial Intelligence to work in your app
- Reply to customers
“Draft a friendly reply to a customer whose parcel is two days late.”
GPT OSS 120B - Translate
“Translate our checkout page into Swahili, keeping the buttons short.”
Nemotron 3 Super - Summarise
“Summarise this support ticket in three lines for the next person on shift.”
GPT OSS 20B
- Describe images
“Write alt text for every product photo in this collection.”
Nemotron 3 Nano Omni - Sort messages
“Tag each message as a question, a complaint, or praise.”
GPT OSS 20B - Read receipts
“Find the shop, the amount, and the date on this receipt.”
Nemotron 3 Nano Omni
- Write product copy
“Write a two-line description for a hand-woven kikoi.”
GPT OSS 120B - Answer from your policy
“Answer this question using only our returns policy.”
GPT OSS 120B - Use your tools
“Check the stock, then tell the customer when their order ships.”
GPT OSS 120B
Models you can use today
- GPT OSS 120Bopenai/gpt-oss-120b
- Made by
- OpenAI
- Reads
- Text
- Takes in
- 128K tokens
- Runs in
- Germany
- GPT OSS 20Bopenai/gpt-oss-20b
- Made by
- OpenAI
- Reads
- Text
- Takes in
- 128K tokens
- Runs in
- Germany
- Nemotron 3 Ultranvidia/nemotron-3-ultra
- Made by
- NVIDIA
- Reads
- Text
- Takes in
- 1M tokens
- Runs in
- United States
- Nemotron 3 Supernvidia/nemotron-3-super
- Made by
- NVIDIA
- Reads
- Text
- Takes in
- 262K tokens
- Runs in
- United States
- Nemotron 3.5 Lightningnvidia/nemotron-3.5-lightning
- Made by
- NVIDIA
- Reads
- Text
- Takes in
- 1M tokens
- Runs in
- United States
- Nemotron 3 Nano Omninvidia/nemotron-3-nano-omni
- Made by
- NVIDIA
- Reads
- Text and images
- Takes in
- 256K tokens
- Runs in
- United States
Everything a real app needs
Works with the OpenAI SDKs
Change the base URL and the key, and the code you already have keeps working, with streaming, images, and tools.
Models that think first
Ask a reasoning model to think harder or answer faster, and choose how much effort each call gets.
Prompts kept in one place
Write a prompt once in the console, version it, and use it from any app in the project by name.
Limits you set
Every project gets a monthly limit, and you can set a lower one, so a busy day never turns into a surprise.
Private by default
We keep a record of each call, never its prompt or its answer, and nothing you send trains a model.
Part of your project
The same keys, team, and bill as your storage, schedules, and webhooks.
Questions about AI
Find answers about cost, privacy, where models run, and limits.
Every call is counted in AI units. Each model has a number of units for every million tokens it reads and writes, and every answer says how many units it used.
Ask your first question
Create an API key in your project, send a message, and see the answer and what it used.