Model

AI21 Jamba 1.5 Mini

OPEN

AI21 Labs · AI21 Jamba family · released Aug 22, 2024

Open hybrid SSM-Transformer, 12B/52B active/total, 256K context under the Jamba Open Model License; strongest Arena-Hard score in its class.

ReasoningCodingVisionFunction callingTool useAgentic

553.6

Elo · rank #289

Parameters

52B

Active params

12B (MoE)

Context

256K tokens

Architecture

Hybrid Mamba-Transformer MoE (Jamba)

License

Jamba Open Model License

Languages

API price (in/out)

$0.2 / $0.4

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

ARC-ChallengeReasoning85.7%#30

best: Llama 3.1 405B · 96.9%

Arena-HardHuman preference46.1%#44

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

BIG-Bench HardReasoning53.4%#92

best: ERNIE 4.5 300B-A47B · 94.3%

BFCLAgents80.7%#11

best: SmolLM3 3B (No Thinking) · 92.3%

GPQA DiamondReasoning32.3%#253

best: GPT-5.6 · 94.6%

GSM8KMath75.8%#91

best: Llama 3.1 405B · 96.8%

HumanEvalCoding62.8%#95

best: Claude Opus 4.5 · 99.4%

IFEvalReasoning75.8%#104

best: Gemma 4 26B A4B · 98.5%

MMLU-ProKnowledge39.8%#152

best: Claude Fable 5 · 91.5%

MMLUKnowledge69.7%#146

best: OpenAI o3 · 92.9%

MMMLUKnowledge64.3%#49

best: Gemini 3.1 Pro · 92.6%

TruthfulQAKnowledge54.1%#41

best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%

WildBenchHuman preference42.4%#7

best: Mistral Small 3.2 (24B) · 65.3%

Run it locally

VRAM @ Q4

29.5 GB

VRAM @ FP16

103.2 GB

Fits on (Q4)

RTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4

AI21 Jamba family

Elo progression across releases

AI21 JambaMar 2024302.4 AI21 Jamba 1.5 LargeAug 2024777.9 AI21 Jamba 1.5 MiniAug 2024553.6

Resources

HFai21labs/AI21-Jamba-Mini-1.5 DOCSdocs.ai21.com APIAvailable via AI21 Labs API

API price $0.2/$0.4 · each benchmark row carries its own source badge (see methodology)