What are the hardware requirements for Shisa 7B V1 AWQ?

Shisa 7B V1 AWQ requires approximately 5.6 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Shisa 7B V1 AWQ and how large is it?

Shisa 7B V1 AWQ is developed by TheBloke, a model with 7b 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 Shisa 7B V1 AWQ?

Shisa 7B V1 AWQ is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Shisa 7B V1 AWQ by TheBloke — VRAM 5.6GB, 32K context

Name: Shisa 7B V1 AWQ
Author: TheBloke

Shisa 7B V1 AWQ is an open-source language model by TheBloke. Features: 7b LLM, VRAM: 5.6GB, Context: 32K, License: apache-2.0, Quantized, LLM Explorer Score: 0.1.

Arxiv:2305.18290 Arxiv:2310.05914 4-bit Awq Base model:augmxnt/shisa-7b-v1 Base model:quantized:augmxnt/s... Conversational Dataset:augmxnt/shisa-en-ja-dp... Dataset:augmxnt/ultra-orca-bor... Dataset:open-orca/slimorca En Ja Mistral Quantized Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/TheBloke/shisa-7B-v1-AWQ

Shisa 7B V1 AWQ Parameters and Internals

Model Type

bilingual, general-purpose, chat, LLM

Use Cases

Areas:

research, commercial applications

Limitations:

higher hallucination rates, less effective instruction following

Considerations:

Additional targeted training or sampler settings like Min P might be needed for language leakage issues.

Supported Languages

english (high), japanese (high)

Training Details

Data Sources:

airoboros-3.1, ultrafeedback_binarized

Methodology:

NEFTune, DPO

Model Architecture:

Custom JA-optimized tokenizer reducing size in Japanese by 50%.

Input Output

Input Format:

[INST] <> {system_message} <> {prompt} [/INST]

Accepted Modalities:

text

Output Format:

text

Performance Tips:

Check Min P settings for better performance

LLM Name	Shisa 7B V1 AWQ
Repository 🤗	https://huggingface.co/TheBloke/shisa-7B-v1-AWQ
Model Name	Shisa 7B v1
Model Creator	AUGMXNT
Base Model(s)	Shisa 7B V1 augmxnt/shisa-7b-v1
Model Size	7b
Required VRAM	5.6 GB
Updated	2026-06-02
Maintainer	TheBloke
Model Type	mistral
Model Files	5.6 GB
Supported Languages	ja en
AWQ Quantization	Yes
Quantization Type	awq
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.35.2
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	120128
Torch Data Type	float16

Rank the Shisa 7B V1 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 Shisa 7B V1 AWQ

Best Alternatives	Context / RAM	Downloads	Likes
Mistral 7B Instruct V0.2 AWQ	32K / 4.2 GB	6	0
Mistral 7B Instruct V0.2 AWQ	32K / 4.2 GB	298182	52
Mistral 7B Instruct V0.3 AWQ	32K / 4.2 GB	5633	10
Mistral 7B Instruct V0.1 AWQ	32K / 4.2 GB	15461	2
Ghost 7B Alpha AWQ	32K / 4.2 GB	14	0
Tess 7B V2.0 AWQ	32K / 4.2 GB	1	1
...reeze 7B 32K Instruct V1.0 AWQ	32K / 4.7 GB	9	0
Breeze 7B Instruct V1.0 AWQ	32K / 4.6 GB	5	0
NSFW DPO Noromaid 7B AWQ	32K / 4.2 GB	206	2
OpenHermes 2.5 Mistral 7B AWQ	32K / 4.2 GB	23691	21

Note: green Score (e.g. "73.2") means that the model is better than TheBloke/shisa-7B-v1-AWQ.