What are the hardware requirements for Codegen 16B Nl?

Codegen 16B Nl requires approximately 32.2 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Codegen 16B Nl and how large is it?

Codegen 16B Nl is developed by Salesforce, a model with 16b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Codegen 16B Nl perform on standard benchmarks?

Codegen 16B Nl has the following published scores: MMLU 32.35. Compare against reference models on this page or on the LLM Explorer leaderboards.

Codegen 16B Nl by Salesforce — VRAM 32.2GB

Name: Codegen 16B Nl
Author: Salesforce

Codegen 16B Nl is an open-source language model by Salesforce. Features: 16b LLM, VRAM: 32.2GB, License: bsd-3-clause, Code Generating, LLM Explorer Score: 0.05, Arc: 46.8, HellaSwag: 71.9, MMLU: 32.4, GSM8K: 2.7.

Arxiv:2203.13474 Codegen Endpoints compatible Pytorch Region:us

Model Card on HF 🤗: https://huggingface.co/Salesforce/codegen-16B-nl

Codegen 16B Nl Parameters and Internals

Model Type

autoregressive, language model

Use Cases

Primary Use Cases:

program synthesis

Limitations:

Best at generating executable code given English prompts

Training Details

Data Sources:

the Pile

Methodology:

Trained using cross-entropy loss for maximizing likelihood.

Hardware Used:

multiple TPU-v4-512

LLM Name	Codegen 16B Nl
Repository 🤗	https://huggingface.co/Salesforce/codegen-16B-nl
Model Size	16b
Required VRAM	32.2 GB
Updated	2026-07-04
Maintainer	Salesforce
Model Type	codegen
Model Files	32.2 GB
Generates Code	Yes
Model Architecture	CodeGenForCausalLM
License	bsd-3-clause
Model Max Length	2048
Transformers Version	4.21.0.dev0
Tokenizer Class	GPT2Tokenizer
Vocabulary Size	51200
Torch Data Type	float16
Activation Function	gelu_new

Rank the Codegen 16B Nl 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 Codegen 16B Nl

Best Alternatives	Context / RAM	Downloads	Likes
Codegen2 16B P	0K / 64.3 GB	1290	45
Instruct Codegen 16B	0K / 32.2 GB	8	21
Codegen 16B Mono Toolbench	0K / 128.4 GB	7	5
Codegen 16B Multi 6 Parts	0K / 32.2 GB	12	0
Codegen 16B Nl Sharded	0K / 32.1 GB	24	7
Fine Tuned Codegen 16B Verilog	0K / 32.2 GB	33	15
Codegen 16B Multi	0K / 32.2 GB	148	119
Codegen 16B Mono	0K / 32.2 GB	89	126

Note: green Score (e.g. "73.2") means that the model is better than Salesforce/codegen-16B-nl.