Skip to content

Jamba2 Mini

Released 2026-01 · chat · 256K tokens · 0 benchmarks

Editorial notes

Lanzado 8 enero 2026. Hybrid SSM-Transformer MoE 12B activos / 52B totales. Mid-trained en 500B tokens curados con math/code y long docs. Liderazgo declarado en IFBench, IFEval, Collie (instruction following) y FACTS (grounding). Apache 2.0. Posicionado como alternativa low-latency a reasoning models (sin "thinking tokens" overhead).

Spec sheet

Company
AI21 Labs
Country
IL
Type
chat
Release
2026-01
Context
256K tokens
Slug
jamba2-mini
No scores published for this model. The editorial notes have additional context.
Cite this model
BibTeX · APA

BibTeX

@misc{frontier-jamba2-mini,
  title  = {Jamba2 Mini},
  author = {{AI21 Labs}},
  year   = {2026},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
  url    = {https://frontierbenchmarks.com/models/jamba2-mini}
}

APA

AI21 Labs (2026). Jamba2 Mini [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/jamba2-mini

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.