# deepseek-ai/DeepSeek-V3.2-Exp

Repository: https://github.com/deepseek-ai/DeepSeek-V3.2-Exp
Canonical: https://ross.abutalabs.com/products/deepseek-v32-exp
Language: Python
License: MIT
License Family: permissive
Last push: 2025-11-18T02:57:11+00:00

## Health v2 (maintenance only)
Score: 40/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 52, release rhythm 35, longevity 24
- inputs: {"age_days": 338, "days_push": 288, "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 1640, forks 191 (observed 2026-08-28T04:05:15.250791+00:00)

## What it is
DeepSeek-V3.2-Exp is an experimental open-weight large language model release that introduces DeepSeek Sparse Attention for more efficient long-context training and inference. The repository provides model weights, inference code, and usage examples for running the model locally.

## Use cases
- run deepseek v3.2 locally
- test sparse attention long-context efficiency
- serve an open-weight LLM with long context
- experiment with efficient transformer attention mechanisms
- download deepseek model weights for inference
- benchmark long-context inference cost savings

## When to choose
- you want to run or evaluate DeepSeek's latest experimental open-weight model
- you're researching sparse attention and long-context efficiency
- you need a permissively licensed (MIT) large language model for self-hosting

## When to avoid
- you need a production-stable model rather than an experimental release
- you lack the GPU hardware to run a very large model
- you only need a hosted chat API without local deployment

## Facets
- artifact type: library
- maturity: experimental
- function: llm-inference, machine-learning, deep-learning, transformers
- domain: large-language-models, deep-learning, artificial-intelligence
- platform: python
- tags: deepseek, sparse-attention, model-weights, long-context, open-weights, transformer-architecture, gpu, linux, docker

## Member repositories
- deepseek-ai/DeepSeek-V3.2-Exp (main) score 40

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:15.250791+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 2026-08-30T03:46:23.726231+00:00, confidence not recorded.
  - readme: https://github.com/deepseek-ai/DeepSeek-V3.2-Exp (fetched 2026-08-28T04:05:15.250791+00:00, sha 12a416bc4999)
- Data as of 2026-08-30T08:39:29.467469+00:00.