# deepseek-ai/DeepSeek-V3

Repository: https://github.com/deepseek-ai/DeepSeek-V3
Canonical: https://ross.abutalabs.com/products/deepseek-v3
Language: Python
License: MIT
License Family: permissive
Last push: 2025-08-28T03:24:37+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 39, release rhythm 8, longevity 43
- inputs: {"age_days": 615, "days_push": 370, "days_rel": 432, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 104416, forks 16733 (observed 2026-08-28T04:12:23.738954+00:00)

## What it is
Official repository for DeepSeek-V3, a 671B-parameter Mixture-of-Experts language model (37B activated per token) with open weights, inference code, and conversion tools. It includes instructions for local deployment via Hugging Face, vLLM, SGLang, and other engines, plus the model's technical report.

## Use cases
- run deepseek v3 locally
- self-host a large open-source LLM
- download deepseek v3 weights from hugging face
- serve a mixture-of-experts model with vllm
- fine-tune or evaluate an open MoE language model
- build a chatbot on an open-weights model

## When to choose
- you need a state-of-the-art open-weights LLM you can self-host
- you have multi-GPU or cluster hardware to run a 671B MoE model
- you want to study or extend a frontier MoE architecture like MLA and multi-token prediction

## When to avoid
- you lack GPUs with hundreds of GB of aggregate VRAM
- you just need a hosted chat API without managing infrastructure
- you need a small model for laptops or edge devices

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, deep-learning
- domain: large-language-models, artificial-intelligence, deep-learning, machine-learning
- platform: python
- tags: mixture-of-experts, open-weights, multi-head-latent-attention, multi-token-prediction, huggingface, self-hosting, gpu, linux, docker

## Member repositories
- deepseek-ai/DeepSeek-V3 (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:12:23.738954+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 2026-08-29T16:12:13.698186+00:00, confidence not recorded.
  - readme: https://github.com/deepseek-ai/DeepSeek-V3 (fetched 2026-08-28T04:12:23.738954+00:00, sha 2d9e3c575705)
- Data as of 2026-08-30T08:39:29.467469+00:00.