# deepseek-ai/DeepSeek-MoE

DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models

Repository: https://github.com/deepseek-ai/DeepSeek-MoE
Canonical: https://ross.abutalabs.com/products/deepseek-moe
Language: Python
License: MIT
License Family: permissive
Last push: 2024-01-16T12:18:10+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 69
- inputs: {"age_days": 974, "days_push": 960, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1969, forks 310 (observed 2026-08-28T04:06:00.705775+00:00)

## What it is
DeepSeekMoE 16B is a Mixture-of-Experts language model with fine-grained expert segmentation and shared expert isolation, released with base and chat checkpoints. The repository provides model code, evaluation details, and quick-start instructions for running the model on a single 40GB GPU.

## Use cases
- run a mixture-of-experts language model on a single GPU
- research MoE expert specialization architectures
- deploy a 16B parameter chat model without quantization
- compare MoE efficiency against dense 7B models
- fine-tune or evaluate an open MoE base model

## When to choose
- you need an open MoE LLM checkpoint for research
- you want to study fine-grained expert segmentation and shared expert isolation
- you have a single 40GB GPU and want a capable chat model

## When to avoid
- you need a maintained production LLM stack with active updates
- you lack a GPU with at least 40GB memory
- you need multilingual support beyond English and Chinese

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, llm-inference, deep-learning
- domain: large-language-models, machine-learning, artificial-intelligence
- platform: python
- tags: mixture-of-experts, model-weights, research-model, huggingface, gpu, linux

## Member repositories
- deepseek-ai/DeepSeek-MoE (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:00.705775+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:05:09.643033+00:00, confidence not recorded.
  - readme: https://github.com/deepseek-ai/DeepSeek-MoE (fetched 2026-08-28T04:06:00.705775+00:00, sha 4bb06d866963)
- Data as of 2026-08-30T08:39:29.467469+00:00.