What are the hardware requirements for Codegen 6B Multi?

Codegen 6B Multi requires approximately 14.3 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Codegen 6B Multi and how large is it?

Codegen 6B Multi is developed by Salesforce, a model with 6b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Codegen 6B Multi perform on standard benchmarks?

Codegen 6B Multi has the following published scores: MMLU 25.71. Compare against reference models on this page or on the LLM Explorer leaderboards.

Codegen 6B Multi by Salesforce — VRAM 14.3GB

Name: Codegen 6B Multi
Author: Salesforce

Codegen 6B Multi is an open-source language model by Salesforce. Features: 6b LLM, VRAM: 14.3GB, License: bsd-3-clause, Code Generating, LLM Explorer Score: 0.06, Arc: 27.2, HellaSwag: 41.1, MMLU: 25.7, GSM8K: 1.

Arxiv:2203.13474 Codegen Deploy:azure Endpoints compatible Pytorch Region:us

Model Card on HF 🤗: https://huggingface.co/Salesforce/codegen-6B-multi

Codegen 6B Multi Parameters and Internals

Model Type

text generation, program synthesis

Use Cases

Areas:

research, commercial applications

Applications:

program synthesis, code generation

Primary Use Cases:

Generating executable code from English prompts, Completing partially generated code

Limitations:

Only best at program synthesis tasks

Considerations:

Prompts should be in the form of a comment string

Supported Languages

Python (High), C (High), C++ (High), Go (High), Java (High), JavaScript (High)

Training Details

Data Sources:

BigQuery, GitHub repositories

Data Volume:

119.2B tokens

Methodology:

Cross-entropy loss to maximize likelihood of sequential inputs

Hardware Used:

TPU-v4-512

Model Architecture:

Autoregressive language model

LLM Name	Codegen 6B Multi
Repository 🤗	https://huggingface.co/Salesforce/codegen-6B-multi
Model Size	6b
Required VRAM	14.3 GB
Updated	2026-06-29
Maintainer	Salesforce
Model Type	codegen
Model Files	14.3 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 6B Multi 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 6B Multi

Best Alternatives	Context / RAM	Downloads	Likes
Nsql 6B	0K / 28.4 GB	23	52
...egen 6B Nl Lora Adapter Merged	0K / 14.3 GB	3	1
...en 6B Mono Lora Adapter Merged	0K / 14.3 GB	5	0
Diff Codegen 6B V2	0K / 14.3 GB	20	40
...en 6B Mono Instruct Py Revised	0K / 28.4 GB	5	2
...n 6B Mono Instruct Py Critique	0K / 28.4 GB	4	2
...gen 6B Nl Instruct Py Critique	0K / 28.4 GB	6	1
Codegen 6B Nl	0K / 14.3 GB	146	4
CodeGen RE	0K / 28.8 GB	5	2
Fine Tuned Codegen 6B Verilog	0K / 28.4 GB	14	2

Note: green Score (e.g. "73.2") means that the model is better than Salesforce/codegen-6B-multi.