# DreamLM/Dream

Dream 7B, a large diffusion language model

Repository: https://github.com/DreamLM/Dream
Canonical: https://ross.abutalabs.com/products/dreamlm-dream
Homepage: https://hkunlp.github.io/blog/2025/dream/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: diffusion-language-models, scalability
Last push: 2025-11-21T09:26:06+00:00

## Health v2 (maintenance only)
Score: 44/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 53, release rhythm 35, longevity 37
- inputs: {"age_days": 518, "days_push": 285, "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 1265, forks 76 (observed 2026-08-28T04:04:10.931933+00:00)

## What it is
Dream 7B is an open diffusion large language model (dLLM) with base and instruct checkpoints, plus inference and training code built on Hugging Face transformers. It matches or exceeds similarly sized autoregressive models on general, math, and coding tasks while offering diffusion-based planning and infilling flexibility.

## Use cases
- run a diffusion language model for text generation
- fine-tune a 7B dLLM on my own dataset
- compare diffusion vs autoregressive LLM performance
- experiment with parallel and infilling text generation
- generate code with an open diffusion model
- research non-autoregressive language model architectures

## When to choose
- you want to experiment with diffusion-based text generation instead of autoregressive decoding
- you need an open-weights 7B model with strong math, coding, and planning ability
- you want training code to fine-tune a dLLM on custom data
- you need variable-length generation or infilling capabilities

## When to avoid
- you need production-ready, widely supported inference like standard autoregressive LLMs
- you lack a GPU with at least 20GB memory
- you need the latest transformers/torch versions, which are untested
- you need multimodal or long-context capabilities beyond this model's scope

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, deep-learning
- domain: large-language-models, artificial-intelligence, deep-learning
- platform: python
- tags: diffusion-language-model, text-generation, huggingface-transformers, 7b-model, open-weights, natural-language-processing, gpu, linux

## Member repositories
- DreamLM/Dream (main) score 44

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:10.931933+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 2026-08-30T05:03:49.853226+00:00, confidence not recorded.
  - readme: https://github.com/DreamLM/Dream (fetched 2026-08-28T04:04:10.931933+00:00, sha d9b4ba0dfae9)
  - homepage: https://hkunlp.github.io/blog/2025/dream/ (fetched 2026-08-29T12:15:59.979047+00:00, sha 9fa09ddc7689)
- Data as of 2026-08-30T08:39:29.467469+00:00.