What are the hardware requirements for Stable Code 3B?

Stable Code 3B requires approximately 5.6 GB of VRAM and supports a context window of 16K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Stable Code 3B and how large is it?

Stable Code 3B is developed by stabilityai, a model with 3b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Stable Code 3B perform on standard benchmarks?

Stable Code 3B has the following published scores: MMLU 37.53, HumanEval (Python) 30.72. Compare against reference models on this page or on the LLM Explorer leaderboards.

Stable Code 3B by stabilityai — VRAM 5.6GB, 16K context

Name: Stable Code 3B
Author: stabilityai

Stable Code 3B is an open-source language model by stabilityai. Features: 3b LLM, VRAM: 5.6GB, Context: 16K, License: other, Quantized, LLM Explorer Score: 0.19, Arc: 35.3, HellaSwag: 57.5, MMLU: 37.5, GSM8K: 19.6, HumanEval: 30.7.

Arxiv:1910.02054 Arxiv:2104.09864 Arxiv:2204.06745 Arxiv:2305.06161 Arxiv:2307.09288 Arxiv:2309.12284 Arxiv:2310.10631 Code Dataset:bigcode/commitpackft Dataset:bigcode/starcoderdata Dataset:bigcode/the-stack-gith... Dataset:eleutherai/proof-pile-... Dataset:meta-math/metamathqa Dataset:tiiuae/falcon-refinedw... En Endpoints compatible Gguf Model-index Q5 Quantized Region:us Safetensors Sharded Stablelm Tensorflow

Model Card on HF 🤗: https://huggingface.co/stabilityai/stable-code-3b

Stable Code 3B Parameters and Internals

Model Type

causal-lm, code, text-generation

Use Cases

Areas:

Research, Commercial applications

Limitations:

The model may exhibit unreliable or undesirable behavior without fine-tuning., Content may be offensive or inappropriate after pre-training.

Considerations:

Evaluation and fine-tuning required prior to deployment for safe performance.

Supported Languages

English (fluent), Code (specialized in 18 programming languages)

Training Details

Data Sources:

tiiuae/falcon-refinedweb, bigcode/the-stack-github-issues, bigcode/commitpackft, bigcode/starcoderdata, EleutherAI/proof-pile-2, meta-math/MetaMathQA

Data Volume:

1.3 trillion tokens

Methodology:

Pre-trained using a mixture of large-scale datasets, applying rotary position embeddings for improved throughput.

Context Length:

16384

Hardware Used:

256 NVIDIA A100 40GB GPUs

Model Architecture:

Decoder-only transformer with modifications similar to LLaMA architecture, utilizing Rotary Position Embeddings.

LLM Name	Stable Code 3B
Repository 🤗	https://huggingface.co/stabilityai/stable-code-3b
Model Size	3b
Required VRAM	5.6 GB
Updated	2026-08-07
Maintainer	stabilityai
Model Type	stablelm
Model Files	5.0 GB: 1-of-2 0.6 GB: 2-of-2 2.0 GB 2.3 GB 5.6 GB
Supported Languages	en
GGUF Quantization	Yes
Quantization Type	gguf\|q5\|q5_k
Model Architecture	StableLmForCausalLM
License	other
Context Length	16384
Model Max Length	16384
Transformers Version	4.38.0
Tokenizer Class	GPTNeoXTokenizer
Vocabulary Size	50304
Torch Data Type	bfloat16

Rank the Stable Code 3B capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Quantized Models of the Stable Code 3B

Model	Likes	Downloads	VRAM
Stable Code 3B GGUF	30	1009	1 GB
Stable Code 3B GPTQ	12	10	1 GB

Best Alternatives to Stable Code 3B

Best Alternatives	Context / RAM	Downloads	Likes
Stable Cypher Instruct 3B	16K / 1.7 GB	864	34
Stable Code Instruct 3B	16K / 5.6 GB	2134	189
Stable Cypher Instruct 3B	16K / 1.7 GB	263	26
...struct 3B Spider 1500 Steps Q4	16K / 1.7 GB	36	0
NSFW 3B	4K / 10.6 GB	15664	194
HELVETE 3B	4K / 10.6 GB	526	199
HelpingAI 3B Chat	4K / 10.6 GB	70	3
Rohit Sharma	4K / 10.6 GB	80	1
Stable Code Instruct 3B 4bit	16K / 1.8 GB	52	5
Stable Code 3B 4bit	16K / 1.8 GB	17	1