LLM
Jamba 1.5 Mini
Jamba 1.5 Mini is an AI21 hybrid SSM-Transformer mixture-of-experts model with a 256K context window, 12B active of 52B total parameters.
Levelmodel
ClassLLM
Parameters52B
Context256,000
Released22 Aug 2024
Opennessopen weights restricted
Runs onmulti-GPU node, single datacentre GPU
Statusnot in the notes
Benchmark Results
vendor-reported
ARC-cknowledge and reasoning
vendor-reported
85.7
050100
Score in %. Higher is better.
Arena-Hardhuman preference arena
vendor-reported
46.1
02550
Score in score. Higher is better.
BFCLagents and tool use
vendor-reported
80.6
050100
Score in %. Higher is better.
GPQAknowledge and reasoning
vendor-reported
32.3
050100
Score in %. Higher is better.
GSM8Kmath
vendor-reported
75.8
050100
Score in %. Higher is better.
MMLUknowledge and reasoning
CoTvendor-reported
69.7
050100
Score in %. Higher is better.
MMLU-Proknowledge and reasoning
CoTvendor-reported
42.5
050100
Score in %. Higher is better.
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
Runtimes
Identity
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- hybrid Hybrid SSM-Transformer (joint attention and Mamba) with mixture of experts
- Total parameters
- 52B sources
- Active parameters
- 12B
- Checkpoint parameters
- –
- Context length
- 256,000 tokens
- Input
- text
- Output
- text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- open weights restricted
- Licence
- Jamba Open Model License Agreement sources
- Access routes
- API, cloud marketplace, open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- English, Spanish, French, Portuguese, Italian, Dutch, German, Arabic, Hebrew
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Lineage
The notes record no lineage for this model.
Industry Use Cases
No industry use case links this model.
Links
DeveloperAI21 Labsstartup, Israel
AI21 Labsstartup, IsraelModelJamba 1.5 MiniAI21 Labs, LLM
Jamba 1.5 MiniAI21 Labs, LLMUse Casesnone recorded
Who Runs Itnone recorded
Organisations 1
Sources
- ai21.com/blog/announcing-jamba-model-family https://www.ai21.com/blog/announcing-jamba-model-family
- huggingface.co/ai21labs/AI21-Jamba-Mini-1.5 https://huggingface.co/ai21labs/AI21-Jamba-Mini-1.5
- huggingface.co/api/models/ai21labs/AI21-Jamba-1.5-Mini https://huggingface.co/api/models/ai21labs/AI21-Jamba-1.5-Mini
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Jamba 1.5 Mini", https://aurion.semai.id/models/ai21-labs-jamba-1-5-mini/, records of October 2026.