Model

StableLM-3B-4e1t

OPEN

Stability AI · StableLM family · released Oct 2, 2023

Strong 3B base trained 4 epochs over ~1T English+code tokens; widely used fine-tuning base (later became Zephyr's foundation).

ReasoningCodingVisionFunction callingTool useAgentic

-203.6

Elo · rank #425

Parameters

2.8B

Active params

2.8B (dense)

Context

4K tokens

Architecture

Dense decoder-only transformer (LLaMA-style, partial RoPE, LayerNorm w/ learned bias)

License

CC BY-SA-4.0

Languages

API price (in/out)

No hosted API

Modalities

text

Benchmark results

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

ARC-ChallengeReasoning46.6%#131

best: Llama 3.1 405B · 96.9%

BIG-Bench HardReasoning9.0%#140

best: ERNIE 4.5 300B-A47B · 94.3%

GPQA DiamondReasoning0.0%#292

best: GPT-5.6 · 94.6%

GSM8KMath3.3%#187

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning75.9%#109

best: Claude 3 Opus · 95.4%

IFEvalReasoning22.0%#150

best: Gemma 4 26B A4B · 98.5%

MATH-500Math1.1%#207

best: GPT-5 · 99.4%

MMLU-ProKnowledge7.4%#185

best: Claude Fable 5 · 91.5%

MMLUKnowledge45.2%#247

best: OpenAI o3 · 92.9%

TruthfulQAKnowledge37.2%#89

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

WinoGrandeReasoning71.2%#94

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

2.2 GB

VRAM @ FP16

5.6 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 · GPTQ

StableLM family

Elo progression across releases

StableLM-Alpha 7BApr 2023-552.2 StableLM-3B-4e1tOct 2023-203.6 StableLM Zephyr 3BDec 2023-0.4

Resources

HFstabilityai/stablelm-3b-4e1t GHStability-AI/StableLM

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