Jamba Tiny Dev
Jamba Tiny Dev is a small trained checkpoint for architecture development and debugging. AI21 reports 319 million parameters and training on about 40 billion tokens.
Research release
A development model rather than a small assistant
The card says the model was not extensively trained and should not be expected to produce high-quality text. Its purpose is to make work on the architecture easier.Jamba-tiny-dev official model card
Useful experiments need a narrow question
Practical guidance. Use a checkpoint like this to inspect loading, training behavior or architectural changes. Do not use its output quality as a measure of the full Jamba family. This trained checkpoint is different from a repository containing randomly initialized weights.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Launch date | |
| Access | Downloadable |
| Model identifiers | ai21labs/Jamba-tiny-dev |
| Parameters | 319 millionJamba-tiny-dev official model card |
| Training volume | About 40 billion tokensJamba-tiny-dev official model card |
| License | Apache 2.0Jamba-tiny-dev official model card |
Related models
Compact dense language model for private document processing and instruction-based tasks.
Launch date Downloadable- Jamba v0.1
AI21 Labs · Jamba
Original pretrained hybrid Mamba-Transformer checkpoint for model training and adaptation.
Launch date Downloadable