What are the hardware requirements for Stablecode Completion Alpha 3B?

Stablecode Completion Alpha 3B requires approximately 14.1 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 Stablecode Completion Alpha 3B and how large is it?

Stablecode Completion Alpha 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 Stablecode Completion Alpha 3B perform on standard benchmarks?

Stablecode Completion Alpha 3B has the following published scores: HumanEval (Python) 20.2. Compare against reference models on this page or on the LLM Explorer leaderboards.

Stablecode Completion Alpha 3B by stabilityai — VRAM 14.1GB, 16K context

Name: Stablecode Completion Alpha 3B
Author: stabilityai

Stablecode Completion Alpha 3B is an open-source language model by stabilityai. Features: 3b LLM, VRAM: 14.1GB, Context: 16K, License: apache-2.0, Code Generating, LLM Explorer Score: 0.08, HumanEval: 20.2.

Arxiv:1910.02054 Arxiv:2104.09864 Code Codegen Dataset:bigcode/starcoderdata Deploy:azure Endpoints compatible Gpt neox Model-index Pytorch Region:us

Model Card on HF 🤗: https://huggingface.co/stabilityai/stablecode-completion-alpha-3b

Stablecode Completion Alpha 3B Parameters and Internals

Model Type

causal-lm

Use Cases

Areas:

code generation

Applications:

single/multiline code completion in multiple programming languages

Primary Use Cases:

code completion from a long context window

Limitations:

Not intended to create unlawful content or engage in activities with high risk of harm.

Additional Notes

The model uses the StarCoder tokenizer with a vocabulary size of 49,000.

Supported Languages

Code (supported)

Training Details

Data Sources:

bigcode/starcoderdata

Data Volume:

300B tokens for pre-training, additional 200B for fine-tuning

Methodology:

Pre-trained using a multi-stage context length extension schedule, first pre-training at a context length of 4096 tokens, then fine-tuning at 16384 tokens.

Context Length:

16384

Model Architecture:

Decoder-only transformer with parallel attention and MLP residuals, single input LayerNorm, rotary position embeddings, LayerNorm bias terms only.

Input Output

Input Format:

Code (programming language syntax)

Accepted Modalities:

text

Output Format:

Code

LLM Name	Stablecode Completion Alpha 3B
Repository 🤗	https://huggingface.co/stabilityai/stablecode-completion-alpha-3b
Model Size	3b
Required VRAM	14.1 GB
Updated	2026-08-04
Maintainer	stabilityai
Model Type	gpt_neox
Model Files	14.1 GB
Supported Languages	code
Generates Code	Yes
Model Architecture	GPTNeoXForCausalLM
License	apache-2.0
Context Length	16384
Model Max Length	16384
Transformers Version	4.29.2
Tokenizer Class	PreTrainedTokenizerFast
Vocabulary Size	49152
Torch Data Type	float16

Rank the Stablecode Completion Alpha 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

Best Alternatives to Stablecode Completion Alpha 3B

Best Alternatives	Context / RAM	Downloads	Likes
...blecode Completion Alpha 3B 4K	4K / 6.1 GB	434	278
Stablecode Instruct Alpha 3B	4K / 6.1 GB	2	304
StableCode 3B	4K / 6.1 GB	8	1
...tion Alpha 3B 4K Openvino Int8	4K / 2.8 GB	8	1
...blecode Instruct Alpha 3B GPTQ	4K / 1.8 GB	10	17
...de Completion Alpha 3B 4K GPTQ	4K / 1.8 GB	6	9
...blecode Instruct Alpha 3B GPTQ	4K / 1.8 GB	5	1

Note: green Score (e.g. "73.2") means that the model is better than stabilityai/stablecode-completion-alpha-3b.