Latest offers SiliconFlow: SiliconFlow new-account $1 free credits View all →
AI21 Labs

AI21 Labs API models & prices

Verified 3 days ago · 2026-10-15

AI21 Labs develops large language models and enterprise AI solutions, including the Jamba family of models, with a focus on efficient inference, long-context processing, and enterprise applications. Current official API purchasing and callability are unverified.

AI Model API Provider · Global

Official website ↗

Maestro is a separate product. Its pricing and any shared trial credits are unverified; these rates apply only to Jamba.

All plans · 2

API models and prices · 4

Usage-based billing: text rates are usually per million tokens; other units and tiers are stated per entry. Compare only the same currency, unit and billing conditions. Unverified cache pricing does not mean free.

Jamba Mini

Official model ID: jamba-mini

Model developer: AI21 Labs · Service provider: AI21 Labs · Market: com

Model family: Jamba · Currency: USD

Rates per million tokens
Price dimensionReference rate / 1M tokens
InputUSD 0.2
OutputUSD 0.4

Official reference rates per million input/output tokens. Current API purchasing and callability are unverified.

Documented API context: 256K · Official rounded display; exact token count unverified

Version / snapshot: 2 / jamba-mini-2-2026-01

Aliases: jamba-mini · jamba-mini-2 · jamba-mini-2-2026-01

Input / output modalities: text → text

Billing provider: AI21 Labs

Use cases: Long-context document processing; Enterprise RAG; Text generation; Information extraction; Enterprise model deployment; API integration

Open weights / self-hosted

ai21labs/AI21-Jamba2-Mini ↗

Requires your own compute and license compliance; does not imply a free hosted API.

Apache-2.0 ↗

Apache-2.0 model weights. Self-hosting requires compute and compliance with the license.

Historical and deprecated models (excluded from current recommendations)
  • jamba-mini-1.7-2025-08 · deprecated · 2026-02-01

    Official snapshot column says 2025-07 while endpoint suffix says 2025-08; preserved verbatim.

  • jamba-large-1.6-2025-03 · deprecated · 2025-08-03
  • jamba-mini-1.6-2025-03 · deprecated · 2025-08-03
  • jamba-large-1.5-2024-08 · deprecated · 2025-05-06
  • jamba-mini-1.5-2024-08 · deprecated · 2025-05-06

Official sources: docs.ai21.com/docs/jamba-foundation-models ↗ · huggingface.co/ai21labs/AI21-Jamba2-Mini ↗

Official prices and API documentation remain public. Studio signup and billing redirect to the marketing website; the unauthenticated models endpoint returns an empty list. New registration, purchase and authenticated inference were not verified. No official whole-service termination notice was verified.

Official source · Last verified: 2026-10-15T08:08:35.221Z

Jamba Large

Official model ID: jamba-large

Model developer: AI21 Labs · Service provider: AI21 Labs · Market: com

Model family: Jamba · Currency: USD

Rates per million tokens
Price dimensionReference rate / 1M tokens
InputUSD 2
OutputUSD 8

Official reference rates per million input/output tokens. Current API purchasing and callability are unverified.

Documented API context: 256K · Official rounded display; exact token count unverified

Version / snapshot: 1.7 / jamba-large-1.7-2025-07

Aliases: jamba-large · jamba-large-1.7 · jamba-large-1.7-2025-07

Input / output modalities: text → text

Billing provider: AI21 Labs

Use cases: Long-context document processing; Enterprise RAG; Text generation; Information extraction; Enterprise model deployment; API integration

Open weights / self-hosted

ai21labs/AI21-Jamba-Large-1.7 ↗

Requires your own compute and license compliance; does not imply a free hosted API.

Jamba Open Model License ↗

Jamba Open Model License; commercial use is restricted above USD 50 million aggregate annual revenue. Gated weights require accepting conditions.

Historical and deprecated models (excluded from current recommendations)
  • jamba-mini-1.7-2025-08 · deprecated · 2026-02-01

    Official snapshot column says 2025-07 while endpoint suffix says 2025-08; preserved verbatim.

  • jamba-large-1.6-2025-03 · deprecated · 2025-08-03
  • jamba-mini-1.6-2025-03 · deprecated · 2025-08-03
  • jamba-large-1.5-2024-08 · deprecated · 2025-05-06
  • jamba-mini-1.5-2024-08 · deprecated · 2025-05-06

Official sources: docs.ai21.com/docs/jamba-foundation-models ↗ · huggingface.co/ai21labs/AI21-Jamba-Large-1.7 ↗

Official prices and API documentation remain public. Studio signup and billing redirect to the marketing website; the unauthenticated models endpoint returns an empty list. New registration, purchase and authenticated inference were not verified. No official whole-service termination notice was verified.

Official source · Last verified: 2026-10-15T08:08:35.221Z

Jamba2 3B

Model developer: AI21 Labs · Service provider: Self-hosted · Market: com

Model family: Jamba · Model catalog only; see service-channel quotes

Open weights for self-hosting; no public official hosted API quote. This does not mean free hosted inference.

Model context: 256K · Official rounded display; exact token count unverified

Input / output modalities: text → text

Use cases: Long-context document processing; Enterprise RAG; Text generation; Information extraction; Enterprise model deployment; API integration

Open weights / self-hosted

ai21labs/AI21-Jamba2-3B ↗

Requires your own compute and license compliance; does not imply a free hosted API.

Apache-2.0 ↗

Apache-2.0 model weights. Self-hosting requires compute and compliance with the license.

Official sources: docs.ai21.com/docs/jamba-foundation-models ↗ · huggingface.co/ai21labs/AI21-Jamba2-3B ↗

Requires your own compute and compliance with the model license; official hosted API availability is not established.

Official source · Last verified: 2026-10-15T08:09:17.5889084Z

Jamba Reasoning 3B

Model developer: AI21 Labs · Service provider: Self-hosted · Market: com

Model family: Jamba · Model catalog only; see service-channel quotes

Open weights for self-hosting; no public official hosted API quote. This does not mean free hosted inference.

Model context: 256K · Official rounded display; exact token count unverified

Input / output modalities: text → text

Use cases: Long-context document processing; Enterprise RAG; Text generation; Information extraction; Enterprise model deployment; API integration

Open weights / self-hosted

ai21labs/AI21-Jamba-Reasoning-3B ↗

Requires your own compute and license compliance; does not imply a free hosted API.

Apache-2.0 ↗

Apache-2.0 model weights. Self-hosting requires compute and compliance with the license.

Official sources: docs.ai21.com/docs/jamba-foundation-models ↗ · huggingface.co/ai21labs/AI21-Jamba-Reasoning-3B ↗

Requires your own compute and compliance with the model license; official hosted API availability is not established.

Official source · Last verified: 2026-10-15T08:09:17.5894357Z

Free Trial — officially advertised terms

USD 10 API trial credits · 7 days validity · No credit card required

Trial credits for new users, separate from model rates; not a permanently free plan.

Redemption and API access are unverified. Reference terms only, excluded from confirmed redeemable offers.

The current pricing page says 7 days; the older usage-cost document says three months. The current pricing page is used without extending the trial. Conflicting source ↗

Current official pricing source ↗ · Verified: 2026-10-15T08:09:20Z

Need a recommendation?

The AI advisor considers your budget, tools, usage and purchase conditions to suggest primary and backup options.

Open AI advisor →