What are the hardware requirements for Llama 2 13B Fp16?

Llama 2 13B Fp16 requires approximately 26 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.

Who developed Llama 2 13B Fp16 and how large is it?

Llama 2 13B Fp16 is developed by TheBloke, a model with 13b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Llama 2 13B Fp16 perform on standard benchmarks?

Llama 2 13B Fp16 has the following published scores: MMLU 55.36. Compare against reference models on this page or on the LLM Explorer leaderboards.

Llama 2 13B Fp16 by TheBloke — VRAM 26GB, 4K context

Name: Llama 2 13B Fp16
Author: TheBloke

Llama 2 13B Fp16 is an open-source language model by TheBloke. Features: 13b LLM, VRAM: 26GB, Context: 4K, Quantized, LLM Explorer Score: 0.11, Arc: 59, HellaSwag: 82.3, MMLU: 55.4, GSM8K: 10.

En Facebook Fp16 Llama Llama2 Meta Pytorch Quantized Region:us Sharded

Model Card on HF 🤗: https://huggingface.co/TheBloke/Llama-2-13B-fp16

Llama 2 13B Fp16 Parameters and Internals

Model Type

text generation

Use Cases

Areas:

Commercial, Research

Primary Use Cases:

Dialogue and natural language generation tasks

Limitations:

Use in languages other than English, Use violating applicable laws

Considerations:

Specific formatting required for chat versions.

Supported Languages

English (primary)

Training Details

Data Sources:

Publicly available online data

Data Volume:

2 trillion tokens

Methodology:

Supervised fine-tuning and reinforcement learning with human feedback (RLHF)

Context Length:

4000

Training Time:

Jan 2023 - July 2023

Hardware Used:

Meta's Research Super Cluster, A100-80GB GPUs

Model Architecture:

Auto-regressive with optimized transformer architecture

Responsible Ai Considerations

Fairness:

Testing has been conducted primarily in English and may not cover all scenarios, hence outputs may be inaccurate or biased.

Transparency:

Model outputs cannot be predicted in advance.

Accountability:

Developers should perform safety testing before deploying applications.

Mitigation Strategies:

Fine-tuning and human feedback alignment strategies used for alignment to human preferences.

Input Output

Input Format:

text

Accepted Modalities:

text

Output Format:

text

Performance Tips:

Follow `INST` and `<>` tags, `BOS` and `EOS` tokens, proper whitespaces for chat models

LLM Name	Llama 2 13B Fp16
Repository 🤗	https://huggingface.co/TheBloke/Llama-2-13B-fp16
Base Model(s)	Ligma L2 13B kubernetes-bad/Ligma-L2-13b
Model Size	13b
Required VRAM	26 GB
Updated	2026-07-08
Maintainer	TheBloke
Model Type	llama
Model Files	9.9 GB: 1-of-3 9.9 GB: 2-of-3 6.2 GB: 3-of-3
Supported Languages	en
Quantization Type	fp16
Model Architecture	LlamaForCausalLM
Context Length	4096
Model Max Length	4096
Transformers Version	4.30.2
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32000
Torch Data Type	float16

Rank the Llama 2 13B Fp16 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

Quantized Models of the Llama 2 13B Fp16

Model	Likes	Downloads	VRAM
LLaMA2 13B Estopia	21	50	26 GB
...MA2 13B Estopia 3.0bpw H6 EXL2	1	3	5 GB
...MA2 13B Estopia 8.0bpw H8 EXL2	1	2	13 GB
Storytelling V2 13B Lora	7	8	0 GB
Lmg V3 Lora	2	4	0 GB
Storytelling V1 13B Lora	7	7	0 GB
Minotaur Llama2 13B Qlora	4	12	1 GB
Minotaur Llama2 13B Qlora	4	3	1 GB

Best Alternatives to Llama 2 13B Fp16

Best Alternatives	Context / RAM	Downloads	Likes
Llama13b 32K Illumeet Finetune	32K / 26 GB	5	0
...Maid V3 13B 32K 8.0bpw H8 EXL2	32K / 13.2 GB	8	1
...Maid V3 13B 32K 6.0bpw H6 EXL2	32K / 10 GB	5	1
WhiteRabbitNeo 13B V1	16K / 26 GB	3146	453
CodeLlama 13B Python Fp16	16K / 26 GB	121	25
CodeLlama 13B Fp16	16K / 26 GB	9	67
CodeLlama 13B Instruct Fp16	16K / 26 GB	137	28
Codellama 13B Bnb 4bit	16K / 7.2 GB	129	5
...Llama 13B Instruct Hf 4bit MLX	16K / 7.8 GB	113	3
WhiteRabbitNeo 13B V1 4bit Mlx	16K / 7.8 GB	147	2

Note: green Score (e.g. "73.2") means that the model is better than TheBloke/Llama-2-13B-fp16.