Stability AI’s StableLM repository contains its language models and related project materials.
Open source
24 alternativesDeepSeek-R1
The DeepSeek-R1 repository links to its reasoning research paper and provides related model materials.
About DeepSeek-R1
This GitHub repository provides materials for DeepSeek-R1, a series of first-generation reasoning language models. It includes a link to the research paper and download links for the model checkpoints. DeepSeek-R1-Zero was trained with large-scale reinforcement learning without supervised fine-tuning as a preliminary step. DeepSeek-R1 adds cold-start data before reinforcement learning to address problems seen in R1-Zero, including endless repetition, poor readability, and language mixing. The model table lists 671B total parameters, 37B activated parameters, and a 128K context length for both R1-Zero and R1.
The repository also offers distilled dense models with 1.5B, 7B, 8B, 14B, 32B, and 70B parameters. These are fine-tuned on samples generated by DeepSeek-R1 and built on Qwen2.5 and Llama 3 base models. The README says the distilled configurations and tokenizers were slightly changed and should be used with the settings provided. Users are asked to review the Usage Recommendation section before running the models locally. The source does not state pricing.
Key features
- Reasoning models DeepSeek-R1-Zero and DeepSeek-R1 trained with reinforcement learning
- Distilled dense models at 1.5B, 7B, 8B, 14B, 32B, and 70B parameters
- Model checkpoints linked for download on Hugging Face
- Listed 128K context length for DeepSeek-R1 and DeepSeek-R1-Zero
- Link to the DeepSeek-R1 research paper and a PDF copy in the repository
- Evaluation settings including a 32,768-token maximum generation length
Best for: Researchers studying reasoning in large language models who want open model weights and documentation for experiments.
Good for
- The models are described as open-sourced, and the repository is open source.
- The authors state that DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.
- Distilled models at several sizes are provided for use with smaller hardware budgets.
Watch out for
- DeepSeek-R1-Zero encounters endless repetition, poor readability, and language mixing.
- Users are asked to review the Usage Recommendation section before running the models locally.
- The distilled models use slightly changed configs and tokenizers and need the settings provided by the authors.
DeepSeek-R1 at a glance
- Category
- AI & Machine Learning › Models & ML Research
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Yes · Source code
- Language
- English
- Last checked
- Oct 6, 2026 · Online when checked
- Listed
- Oct 5, 2026
DeepSeek-R1 alternatives
Tools and sites you could use instead, ranked by how closely they match.
This repository contains the DeepSeek-V3 language model project and related source files.
Open source
24 alternatives
Repository accompanying the paper Generative Agents: Interactive Simulacra of Human Behavior, containing the core simulation module for agents that simulate believable human behaviors and their game environment.
Open source
24 alternatives
This repository provides a PyTorch implementation of a CVPR paper on high resolution representations for human pose estimation.
Open source
24 alternatives
This repository contains Google Research code for the Flan-T5 language model project.
Open source
24 alternatives
The S4 repository presents structured state space sequence models for sequence modeling research.
Open source
24 alternativesPopular in Models & ML Research
Well-known listings from the same topic.
Falcon LLMfalconllm.tii.aeERNIEernie.baidu.comTokenVersetoken-verse.github.iowTAObittensor.comAlpacacrfm.stanford.edu/2023/03/13/alpaca.htmlSmolVLAhuggingface.co/blog/smolvla
DeepSeek-R1 FAQ
Does the repository say what to check before running the models locally?
Yes. The README recommends reviewing the Usage Recommendation section before running the DeepSeek-R1 series models locally.
What base models are the distilled versions built on?
The distilled models are based on Qwen2.5-Math-1.5B, Qwen2.5-Math-7B, Llama-3.1-8B, Qwen2.5-14B, Qwen2.5-32B, and Llama-3.3-70B-Instruct.
What context length is listed for the main models?
The model table lists a 128K context length for both DeepSeek-R1-Zero and DeepSeek-R1.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.