What are the hardware requirements for Codellama 70B Instruct Nf4 Fp16 Upscaled?

Codellama 70B Instruct Nf4 Fp16 Upscaled requires approximately 138.7 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 Codellama 70B Instruct Nf4 Fp16 Upscaled and how large is it?

Codellama 70B Instruct Nf4 Fp16 Upscaled is developed by arnavgrg, a model with 70b 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 Codellama 70B Instruct Nf4 Fp16 Upscaled?

Codellama 70B Instruct Nf4 Fp16 Upscaled is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Codellama 70B Instruct Nf4 Fp16 Upscaled by arnavgrg — VRAM 138.7GB, 4K context

Name: Codellama 70B Instruct Nf4 Fp16 Upscaled
Author: arnavgrg

Codellama 70B Instruct Nf4 Fp16 Upscaled is an open-source language model by arnavgrg. Features: 70b LLM, VRAM: 138.7GB, Context: 4K, License: apache-2.0, Quantized, Instruction-Based, Code Generating, LLM Explorer Score: 0.11.

Codegen Conversational Endpoints compatible Fp16 Instruct Llama Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/arnavgrg/codellama-70b-instruct-nf4-fp16-upscaled

Codellama 70B Instruct Nf4 Fp16 Upscaled Parameters and Internals

Model Type

text generation

Additional Notes

Quantization to nf4 is not lossless, and model weights for linear layers are lossy compared to the official base model.

Training Details

Methodology:

Upscaled fp16 variant after nf4 4-bit quantization

Input Output

Accepted Modalities:

text

Performance Tips:

Upscaling helps avoid quantization/dequantization costs for each inference pass.

LLM Name	Codellama 70B Instruct Nf4 Fp16 Upscaled
Repository 🤗	https://huggingface.co/arnavgrg/codellama-70b-instruct-nf4-fp16-upscaled
Model Size	70b
Required VRAM	138.7 GB
Updated	2026-08-02
Maintainer	arnavgrg
Model Type	llama
Instruction-Based	Yes
Model Files	4.7 GB: 1-of-29 4.7 GB: 2-of-29 5.0 GB: 3-of-29 5.0 GB: 4-of-29 4.7 GB: 5-of-29 4.7 GB: 6-of-29 4.7 GB: 7-of-29 5.0 GB: 8-of-29 5.0 GB: 9-of-29 4.7 GB: 10-of-29 4.7 GB: 11-of-29 4.7 GB: 12-of-29 5.0 GB: 13-of-29 5.0 GB: 14-of-29 4.7 GB: 15-of-29 4.7 GB: 16-of-29 4.7 GB: 17-of-29 5.0 GB: 18-of-29 5.0 GB: 19-of-29 4.7 GB: 20-of-29 4.7 GB: 21-of-29 4.7 GB: 22-of-29 5.0 GB: 23-of-29 5.0 GB: 24-of-29 4.7 GB: 25-of-29 4.7 GB: 26-of-29 4.7 GB: 27-of-29 5.0 GB: 28-of-29 3.8 GB: 29-of-29
Quantization Type	fp16
Generates Code	Yes
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	4096
Model Max Length	4096
Transformers Version	4.37.0
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32016
Torch Data Type	float16

Rank the Codellama 70B Instruct Nf4 Fp16 Upscaled 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 Codellama 70B Instruct Nf4 Fp16 Upscaled

Best Alternatives	Context / RAM	Downloads	Likes
...Llama 70B Instruct Hf 4bit MLX	4K / 39.1 GB	417	25
...70B Instruct Hf 5.0bpw H6 EXL2	2K / 43.6 GB	5	6
...70B Instruct Hf 2.4bpw H6 EXL2	2K / 21.3 GB	6	1
...0B Instruct Hf 2.65bpw H6 EXL2	2K / 23.4 GB	5	3
...70B Instruct Hf 4.0bpw H6 EXL2	2K / 35.1 GB	6	1
CodeLlama 70B Instruct Hf	4K / 72.3 GB	729	24
Code Llama 70B Python Instruct	4K / 138.1 GB	5	1
CodeLlama 70B Instruct Hf	4K / 72.3 GB	463	210
CodeLlama 70B Instruct Hf GGUF	4K / 25.5 GB	487	2
CodeLlama 70B Instruct AWQ	4K / 36.6 GB	297	13

Note: green Score (e.g. "73.2") means that the model is better than arnavgrg/codellama-70b-instruct-nf4-fp16-upscaled.