Model

Stable LM 2 1.6B

OPEN

Stability AI · StableLM 2 family · released Jan 19, 2024

Compact 7-language multilingual model (English, Spanish, German, Italian, French, Portuguese, Dutch) trained on 2T tokens over 2 epochs; smallest member of the StableLM 2 line.

ReasoningCodingVisionFunction callingTool useAgentic

-229.9

Elo · rank #428

Parameters

1.64B

Active params

1.64B (dense)

Context

4K tokens

Architecture

Dense decoder-only transformer (24 layers, LLaMA-style partial RoPE + LayerNorm, Arcade100k tokenizer)

License

Stability AI Community License (non-commercial by default; commercial use requires Stability AI membership)

Languages

API price (in/out)

No hosted API

Modalities

text

Benchmark results

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

ARC-ChallengeReasoning43.7%#142

best: Llama 3.1 405B · 96.9%

BIG-Bench HardReasoning8.6%#141

best: ERNIE 4.5 300B-A47B · 94.3%

GPQA DiamondReasoning0.0%#291

best: GPT-5.6 · 94.6%

GSM8KMath35.2%#148

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning69.3%#127

best: Claude 3 Opus · 95.4%

IFEvalReasoning11.6%#156

best: Gemma 4 26B A4B · 98.5%

MATH-500Math0.8%#208

best: GPT-5 · 99.4%

MMLU-ProKnowledge5.2%#186

best: Claude Fable 5 · 91.5%

MMLUKnowledge41.9%#254

best: OpenAI o3 · 92.9%

MT-BenchHuman preference5.42#67

best: Hunyuan-Large (A52B) · 9.4

TruthfulQAKnowledge45.2%#65

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

WinoGrandeReasoning64.1%#127

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

1.1 GB

VRAM @ FP16

3.3 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF

StableLM 2 family

Elo progression across releases

Stable LM 2 1.6BJan 2024-229.9 Stable LM 2 12BApr 2024440.3

Resources

HFstabilityai/stablelm-2-1_6b GHStability-AI/StableLM

API price weights · each benchmark row carries its own source badge (see methodology)