What are the hardware requirements for ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized?

ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized requires approximately 7.8 GB of VRAM and supports a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Where can I download or evaluate ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized?

ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized by SushiTokyo — VRAM 7.8GB, 4K context

Name: ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized
Author: SushiTokyo

ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized is an open-source language model by SushiTokyo. Features: 13b LLM, VRAM: 7.8GB, Context: 4K, Quantized, Instruction-Based, LLM Explorer Score: 0.12.

4-bit 4bit Endpoints compatible Gptq Instruct Llama Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/SushiTokyo/ELYZA-japanese-Llama-2-13b-fast-instruct-4bit-quantized

ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized Parameters and Internals

Model Type

text generation, instruction-following

Additional Notes

The model is quantized to 4-bit for faster processing. Original model faced difficulty even with RTX4090, but now outputs in about 10 seconds. Accuracy is unmeasured. Refer to quantize.py for quantization code.

LLM Name	ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized
Repository 🤗	https://huggingface.co/SushiTokyo/ELYZA-japanese-Llama-2-13b-fast-instruct-4bit-quantized
Model Size	13b
Required VRAM	7.8 GB
Updated	2026-08-06
Maintainer	SushiTokyo
Model Type	llama
Instruction-Based	Yes
Model Files	5.0 GB: 1-of-2 2.8 GB: 2-of-2
Quantization Type	4bit
Model Architecture	LlamaForCausalLM
Context Length	4096
Model Max Length	4096
Transformers Version	4.40.1
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	44581
Torch Data Type	float16

Rank the ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized 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 ELYZA Japanese Llama 2 13B Fast Instruct 4bit Quantized

Best Alternatives	Context / RAM	Downloads	Likes
CodeLlama 13B Instruct Fp16	16K / 26 GB	137	28
...Llama 13B Instruct Hf 4bit MLX	16K / 7.8 GB	113	3
...13B Instruct Nf4 Fp16 Upscaled	16K / 26 GB	6	0
Model 007 13b V2	4K / 26 GB	57	4
...igogne2 Enno 13B Sft Lora 4bit	4K / 26 GB	786	0
Xwin LM 13B V0.2 EXL2	4K / 5.2 GB	12	3
Mythalion 13B 2.30bpw H4 EXL2	4K / 4.1 GB	4	3
...lion Kimiko V2.6.05bpw H8 EXL2	4K / 10.1 GB	5	1
Finance LLM 13B 6.0bpw H6 EXL2	2K / 10 GB	5	1
Law LLM 13B 4.0bpw H6 EXL2	2K / 6.8 GB	2	1

Note: green Score (e.g. "73.2") means that the model is better than SushiTokyo/ELYZA-japanese-Llama-2-13b-fast-instruct-4bit-quantized.