← All providers
AI21 Labs pricing
verified 2026-09-02Maker of the Jamba models — hybrid SSM-transformer architecture with long context at mid-tier prices.
data residency & retention
from AI21 Labs's own policies · verified 2026-09-02 · some facts not publishedHeadquarters
🇮🇱 Israel (Other)
Where inference runs
Not disclosed
AI21 does not disclose where Studio/Jamba API inference runs; its privacy policy says personal data may be processed in Israel, the US, the EEA, the UK, and other locations.
Region pinning
Not disclosed
No region-pinning is disclosed for AI21's own API; region control is only achievable indirectly via third-party clouds (Bedrock, Vertex).
Prompt retention
Not disclosed
No fixed retention window is published; the privacy policy states only that data is retained as long as necessary, then deleted or anonymized.
Trains on API data
No
AI21 does not use customer inputs or outputs to train its generally available models unless otherwise agreed in writing (opt-in).
Zero data retention
Not disclosed
Not disclosed on AI21's public pages; third-party profiles reference an optional 'traceless mode', but AI21's own docs do not document it.
current models · $ per 1M tokens
| Model | Input /1M | Cached in | Output /1M | Context |
|---|---|---|---|---|
Jamba Large Jamba | $2 | — | $8 | 256K |
Jamba Mini Jamba | $0.20 | — | $0.40 | 256K |
AI21 Labs totals · 1M in + 1M out
Jamba Mini
$0.6000
Jamba Large
$10.00
Cost of 1M input + 1M output tokens. Bar length is square-root scaled so cheap models stay visible next to premium ones.