What are the hardware requirements for Starcoder2 3B?

Starcoder2 3B requires approximately 12.1 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 Starcoder2 3B and how large is it?

Starcoder2 3B is developed by bigcode, 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 Starcoder2 3B perform on standard benchmarks?

Starcoder2 3B has the following published scores: MMLU 38.65, HumanEval (Python) 31.44. Compare against reference models on this page or on the LLM Explorer leaderboards.

Starcoder2 3B by bigcode — VRAM 12.1GB, 16K context

Name: Starcoder2 3B
Author: bigcode

Starcoder2 3B is an open-source language model by bigcode. Features: 3b LLM, VRAM: 12.1GB, Context: 16K, License: bigcode-openrail-m, LLM Explorer Score: 0.24, Arc: 34.6, HellaSwag: 47.6, MMLU: 38.7, GSM8K: 19.6, HumanEval: 31.4.

Arxiv:2004.05150 Arxiv:2205.14135 Arxiv:2207.14255 Arxiv:2305.13245 Arxiv:2402.19173 Code Dataset:bigcode/the-stack-v2-t... Deploy:azure Endpoints compatible Model-index Region:us Safetensors Starcoder2

Model Card on HF 🤗: https://huggingface.co/bigcode/starcoder2-3b

Starcoder2 3B Benchmarks

ARC: 34.56 vs 96.7 (so35)^-64.3%

HellaSwag: 47.62 vs 95.3 (gpt4)^-50%

MMLU: 38.65 vs 88.3 (so35)^-56.2%

GSM8K: 19.64 vs 96.4 (so35)^-79.6%

HumanEvalA: 23.43

Python Eva: 31.44 vs 92 (so35)^-65.8%

Java Eval: 27.41

Javascript: 35.37

CPP Eval: 27.24

PHP Eval: 27.61

Julia Eval: 19.87

Swift Eval: 25.09

LLME Score: 0.23628

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

🌟 Advertise your project 🚀

Starcoder2 3B Parameters and Internals

Model Type

text-generation

Use Cases

Areas:

research, programming

Primary Use Cases:

code generation

Limitations:

Generated code may contain bugs or exploits., Inefficiency and unintended outputs

Supported Languages

programming_languages (17 programming languages,)

Training Details

Data Sources:

GitHub code, Arxiv, Wikipedia

Data Volume:

3+ trillion tokens

Methodology:

Grouped Query Attention, Fill-in-the-Middle objective

Context Length:

16384

Hardware Used:

160 A100 GPUs

Model Architecture:

Transformer decoder with grouped-query and sliding window attention

Input Output

Accepted Modalities:

text

LLM Name	Starcoder2 3B
Repository 🤗	https://huggingface.co/bigcode/starcoder2-3b
Model Size	3b
Required VRAM	12.1 GB
Updated	2026-05-17
Maintainer	bigcode
Model Type	starcoder2
Model Files	12.1 GB
Model Architecture	Starcoder2ForCausalLM
License	bigcode-openrail-m
Context Length	16384
Model Max Length	16384
Transformers Version	4.37.0.dev0
Tokenizer Class	GPT2Tokenizer
Vocabulary Size	49152

Quantized Models of the Starcoder2 3B

Model	Likes	Downloads	VRAM
Starcoder2 3B AWQ	0	17	2 GB
StarCoder2 3B GGUF	19	2268	1 GB

Best Alternatives to Starcoder2 3B

Best Alternatives	Context / RAM	Downloads	Likes
Starcoder2 3b AutoRedteam	16K / 12.7 GB	5	0
Starcoder Proto Code	16K / 6.1 GB	5	0
Mojo Starcoder2	16K / 6.4 GB	11	0
Starcoder2 3B Instruct	16K / 6.1 GB	1764	5
NEARCoder 3B	16K / 6.1 GB	5	0
NEAR StructTunedStarcoder2	16K / 6.1 GB	8	0
NEAR PreTrainedStarCoder2	16K / 6.1 GB	5	0
Bigcode Starcoder2 3B 8bits	16K / 3.2 GB	5	0
OpenCodeInterpreter SC2 3B	16K / 6.4 GB	22	7
Opencsg Starcoder2 3B V0.1	16K / 6.4 GB	6	1

Rank the Starcoder2 3B Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

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

What open-source LLMs or SLMs are you in search of? 53999 in total.

Email us: info@extractum.io. Our Privacy Policy | Terms and Conditions | Suggest an improvement.

Our Social Media →

Original data from HuggingFace, OpenCompass and various public git repos.

Check out Ag3ntum — our secure, self-hosted AI agent for server management.

Release v20260328a

Support LLM Explorer

Starcoder2 3B by bigcode