AI21 Labs

Jamba, the hybrid Mamba-Transformer that stays cheap on very long documents.

AI21 Labs is the Israeli lab behind Jamba, a hybrid Mamba-Transformer that stays cheap on long documents because it does not pay full Transformer cost on every token. 256K of context. New accounts get $10 of credits for three months.

Two models we track, OpenAI-compatible, no card to start. The old /studio/v1/models listing now returns 410 — pick the current Jamba id from the console. Use it when the document is the point, not when you want a general chat model.

Free models: 2
Max context: 256K
Free tier: $10 of credits for 3 months
Requirement: Email account, no card

Endpoint

Base URLhttps://api.ai21.com/studio/v1

Env varAI21_API_KEY

Rate limit$10 of credits for 3 months

OpenAI-compatible
Native SDK: AI21 SDK
text

Code examples

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.ai21.com/studio/v1',
  apiKey: process.env.AI21_API_KEY
})

const response = await client.chat.completions.create({
  model: 'jamba-mini-2',
  messages: [{ role: 'user', content: 'Explain closures in one paragraph.' }]
})

console.log(response.choices[0].message.content)

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.ai21.com/studio/v1",
    api_key=os.environ["AI21_API_KEY"],
)

response = client.chat.completions.create(
    model="jamba-mini-2",
    messages=[{"role": "user", "content": "Explain closures in one paragraph."}],
)

print(response.choices[0].message.content)

curl https://api.ai21.com/studio/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AI21_API_KEY" \
  -d '{
    "model": "jamba-mini-2",
    "messages": [{"role": "user", "content": "Explain closures in one paragraph."}]
  }'

pip install ai21

import os
from ai21 import AI21Client
from ai21.models.chat import ChatMessage

client = AI21Client(api_key=os.environ["AI21_API_KEY"])

response = client.chat.completions.create(
    model="jamba-mini-2",
    messages=[ChatMessage(role="user", content="Explain closures in one paragraph.")],
)

print(response.choices[0].message.content)

Worth knowing

The /studio/v1/models listing now returns 410. Use the current Jamba ids from the console.

Models you get for free2

Jamba Mini 2jamba-mini-2
Jamba Large 1.6jamba-large-1.6