AI21 Labs
Efficient

Jamba 1.5 Mini

Efficient

Smaller Jamba 1.5, same SSM/Transformer hybrid, single-GPU friendly. Aimed at cheap long-context inference.

Jamba 1.5 Mini is a efficient AI model from AI21 Labs. It costs $0.2000 per million input tokens and $0.4000 per million output tokens (blended $0.2600/M), with a 256,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$0.2000/M
per million input tokens
OUTPUT
$0.4000/M
per million output tokens
BLENDED 70/30
$0.2600/M
unchanged since 3 May
CONTEXT
256,000
tokens

Hand-maintained from AI21 Labs's published pricing; no check date on record.

What it is good at
  • Single-GPU
  • Long context economics
  • Open weights (Jamba license)
Typical use cases
  • Long-document RAG on a budget
  • On-prem long context

Benchmarks

vs. best public score
MMLU69.7%
Multitask academic knowledge across 57 subjects. Source: AI21 Labs (self-reported), MMLU (5-shot, Jamba-1.5-Mini), read 2026-11-10.
HumanEval62.8%
Python function synthesis from docstrings. Source: AI21 Labs (self-reported), HumanEval, read 2026-11-10.
LMArena Elo1240 Elo
Crowd-sourced head-to-head preference Elo rating. Source: LMArena (independent), LMArena Text Arena score, read 2026-11-10.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from AI21 Labs

See all 5 →

Frequently asked questions

How much does Jamba 1.5 Mini cost?

Jamba 1.5 Mini costs $0.2000 per million input tokens and $0.4000 per million output tokens, for a blended reference rate of $0.2600 per million tokens.

What is Jamba 1.5 Mini's context window?

Jamba 1.5 Mini supports up to 256,000 tokens of context in a single request.

What is Jamba 1.5 Mini best for?

Jamba 1.5 Mini is well suited to Single-GPU, Long context economics and Open weights (Jamba license).

Who makes Jamba 1.5 Mini?

Jamba 1.5 Mini is developed and served by AI21 Labs. It was released in Aug 2024.

Terms used on this page