What are the hardware requirements for Starchat2 15B V0.1 AWQ?

Starchat2 15B V0.1 AWQ requires approximately 9.2 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 Starchat2 15B V0.1 AWQ and how large is it?

Starchat2 15B V0.1 AWQ is developed by stelterlab, a model with 15b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Starchat2 15B V0.1 AWQ?

Starchat2 15B V0.1 AWQ is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Starchat2 15B V0.1 AWQ by stelterlab — VRAM 9.2GB, 16K context

Name: Starchat2 15B V0.1 AWQ
Author: stelterlab

Starchat2 15B V0.1 AWQ is an open-source language model by stelterlab. Features: 15b LLM, VRAM: 9.2GB, Context: 16K, License: bigcode-openrail-m, Quantized, LLM Explorer Score: 0.12.

Arxiv:2311.07911 Arxiv:2402.19173 4-bit Awq Base model:huggingfaceh4/starc... Base model:quantized:huggingfa... Conversational Dataset:huggingfaceh4/orca dpo... Dataset:huggingfaceh4/ultrafee... Endpoints compatible Quantized Region:us Safetensors Starcoder2

Model Card on HF 🤗: https://huggingface.co/stelterlab/starchat2-15b-v0.1-AWQ

Starchat2 15B V0.1 AWQ Parameters and Internals

Model Type

GPT-like, fine-tuned

Use Cases

Areas:

chat, programming

Applications:

coding assistant

Primary Use Cases:

code completion, programming help

Limitations:

syntactically valid but semantically incorrect code possible, demographic bias aligned with GitHub community, false URLs

Considerations:

Review output for accuracy and potential biases.

Additional Notes

Model not aligned with RLHF, may produce unaligned outputs.

Supported Languages

English (primary), programming_languages (600+)

Training Details

Data Sources:

HuggingFaceH4/ultrafeedback_binarized, HuggingFaceH4/orca_dpo_pairs

Methodology:

Fine-tuned from StarCoder2, using SFT and DPO techniques.

Hardware Used:

multi-GPU, 8 devices

Model Architecture:

GPT-like

Input Output

Input Format:

Messages in chat format

Accepted Modalities:

text

Output Format:

Generated text responses

Performance Tips:

Utilize pipelines for optimal performance.

LLM Name	Starchat2 15B V0.1 AWQ
Repository 🤗	https://huggingface.co/stelterlab/starchat2-15b-v0.1-AWQ
Base Model(s)	HuggingFaceH4/starchat2-15b-sft-v0.1 HuggingFaceH4/starchat2-15b-sft-v0.1
Model Size	15b
Required VRAM	9.2 GB
Updated	2026-07-29
Maintainer	stelterlab
Model Type	starcoder2
Model Files	9.2 GB
AWQ Quantization	Yes
Quantization Type	awq
Model Architecture	Starcoder2ForCausalLM
License	bigcode-openrail-m
Context Length	16384
Model Max Length	16384
Transformers Version	4.40.1
Tokenizer Class	GPT2Tokenizer
Padding Token	<\|im_end\|>
Vocabulary Size	49154
Torch Data Type	float16

Rank the Starchat2 15B V0.1 AWQ 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 Starchat2 15B V0.1 AWQ

Best Alternatives	Context / RAM	Downloads	Likes
Starcoder2 15B AWQ	16K / 9.2 GB	8	1
Speechless Starcoder2 15B	16K / 31.9 GB	9	3
Starcoder2 15B	16K / 63.8 GB	8388	674
Starcoder2 15B Instruct V0.1	16K / 31.9 GB	597	105
...rCoder2 15B Instruct V0.1 GGUF	16K / 6.2 GB	412	0
StarCoder2 15B GGUF	16K / 6.2 GB	1731	31
...rCoder2 15B Instruct V0.1 GGUF	16K / 6.2 GB	261	0
Starcoder2 15B Finetuned Drake	16K / 63.8 GB	7	0
CodeFuse StarCoder2 15B	16K / 31.9 GB	18	1
Starchat2 15B V0.1	16K / 31.9 GB	352	112

Note: green Score (e.g. "73.2") means that the model is better than stelterlab/starchat2-15b-v0.1-AWQ.