What are the hardware requirements for Jetmoe 8B Sft?

Jetmoe 8B Sft requires approximately 17 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Jetmoe 8B Sft and how large is it?

Jetmoe 8B Sft is developed by jetmoe, a model with 8b 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 Jetmoe 8B Sft?

Jetmoe 8B Sft is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Jetmoe 8B Sft by jetmoe — VRAM 17GB

Name: Jetmoe 8B Sft
Author: jetmoe

Jetmoe 8B Sft is an open-source language model by jetmoe. Features: 8b LLM, VRAM: 17GB, License: apache-2.0, LLM Explorer Score: 0.11.

Arxiv:2404.07413 Alignment-handbook Base model:finetune:jetmoe/jet... Base model:jetmoe/jetmoe-8b Conversational Dataset:huggingfaceh4/airoboro... Dataset:huggingfaceh4/capybara Dataset:huggingfaceh4/code-fee... Dataset:huggingfaceh4/orca-mat... Dataset:huggingfaceh4/systemch... Dataset:huggingfaceh4/ultracha... Endpoints compatible Generated from trainer Jetmoe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/jetmoe/jetmoe-8b-sft

Jetmoe 8B Sft Parameters and Internals

Model Type

Causal Language Model

Training Details

Data Sources:

HuggingFaceH4/ultrachat_200k, HuggingFaceH4/airoboros-3.2, HuggingFaceH4/Code-Feedback, HuggingFaceH4/orca-math-word-problems-200k, HuggingFaceH4/SystemChat, HuggingFaceH4/capybara

Data Volume:

1.25 trillion tokens

Methodology:

Two-phases training method from MiniCPM. Phase 1 uses a constant learning rate with linear warmup and is trained on 1 trillion tokens from large-scale open-source pretraining datasets, including RefinedWeb, Pile, Github data, etc. Phase 2 uses exponential learning rate decay and is trained on 250 billion tokens from phase 1 datasets and extra high-quality open-source datasets.

Hardware Used:

96×H100 GPU cluster

Model Architecture:

24 blocks with two MoE layers per block: Mixture of Attention heads (MoA) and Mixture of MLP Experts (MoE). Each MoA and MoE layer has 8 experts with 2 activated per input token. 8 billion total parameters and 2.2B active during inference.

LLM Name	Jetmoe 8B Sft
Repository 🤗	https://huggingface.co/jetmoe/jetmoe-8b-sft
Base Model(s)	Jetmoe 8B jetmoe/jetmoe-8b
Model Size	8b
Required VRAM	17 GB
Updated	2026-07-30
Maintainer	jetmoe
Model Type	jetmoe
Model Files	4.9 GB: 1-of-4 4.9 GB: 2-of-4 4.9 GB: 3-of-4 2.3 GB: 4-of-4
Model Architecture	JetMoEForCausalLM
License	apache-2.0
Model Max Length	4096
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32000
Activation Function	silu

Rank the Jetmoe 8B Sft 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 Jetmoe 8B Sft

Best Alternatives	Context / RAM	Downloads	Likes
Jetmoe 8B	0K / 17 GB	3917	253
Jetmoe 8B Chat	0K / 17 GB	37	32

Note: green Score (e.g. "73.2") means that the model is better than jetmoe/jetmoe-8b-sft.