Jamba Reasoning 3B

AI21 Labs

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0 / $0

in / out

Context

—

Avg score

20.9

10 benchmarks

AA Intelligence

4.1

index

Weights

Closed

Released

Oct 8, 2025

About

Jamba Reasoning 3B is a proprietary model in the Jamba family from AI21 Labs. Benchmarked on 10 evals, averaging 20.9.

Benchmark scores · 10

avg 20.9 — higher is better

MMLU-Proartificial-analysis · 2026-07

57.7

IFBenchartificial-analysis · 2026-07

52.4

GPQA Diamondartificial-analysis · 2026-07

33.3

LiveCodeBenchartificial-analysis · 2026-07

τ²-benchartificial-analysis · 2026-07

15.8

AIMEartificial-analysis · 2026-07

10.7

AA Long-Context Reasoningartificial-analysis · 2026-07

SciCodeartificial-analysis · 2026-07

5.9

Humanity's Last Examartificial-analysis · 2026-07

4.6

Terminal-Bench Hardartificial-analysis · 2026-07

0.8

Details

PublisherAI21 Labs

FamilyJamba

Parameters—

License—

WeightsClosed

ReleasedOct 8, 2025

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0/M