Skip to content
TrustList

TrustList summary

The DeepSeek-R1 repository, published by DeepSeek on GitHub under an MIT licence, introduces the company's first generation of reasoning models: DeepSeek-R1-Zero and DeepSeek-R1. It is where the weights, the technical paper and the usage guidance are collected.

Reinforcement learning without a supervised start

DeepSeek-R1-Zero was trained with large-scale reinforcement learning applied directly to a base model, with no supervised fine-tuning beforehand. DeepSeek reports that this produced behaviours such as self-verification, reflection and long chains of thought, and calls it the first open research to validate that reasoning ability can be developed through reinforcement learning alone. The same text lists its weaknesses: endless repetition, poor readability and mixed languages.

How R1 addresses those weaknesses

DeepSeek-R1 adds cold-start data before reinforcement learning. The pipeline uses two reinforcement learning stages, aimed at better reasoning patterns and at alignment with human preferences, and two supervised stages that seed reasoning and non-reasoning abilities. DeepSeek reports performance comparable to OpenAI-o1 on maths, code and reasoning tasks.

Sizes, context and distilled models

Both R1-Zero and R1 are built on DeepSeek-V3-Base, with 671B total parameters, 37B activated per token and a 128K context length. DeepSeek also open-sourced six dense models distilled from R1 on Qwen2.5 and Llama3 bases, at 1.5B, 7B, 8B, 14B, 32B and 70B. It reports that the 32B Qwen distillation outperforms OpenAI-o1-mini on various benchmarks, and it asks readers to review a usage recommendation section before running the models locally.

Summary written by TrustList from the business's own published information, Nov 2026.

Benchmarks & AI stats

BenchmarkOfficialCommunity avg
GPQA Diamondreasoning71.5 %-
MATH-500reasoning97.3 %-
MMLUllm90.8 %-

“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).

Sign in to add a benchmark score for this model.

DeepSeek-R1 launches

Request a demo or quote from DeepSeek-R1

Protected by reCAPTCHA - Google Privacy Policy and Terms apply.

By sending, you agree we may share your request and contact details with the provider once you confirm your email.

Reviews

Write the first review of DeepSeek-R1

Used it? Your experience helps other buyers decide.

Write a review

Questions & answers

No questions yet. Be the first to ask about DeepSeek-R1.

DeepSeek-R1 alternatives

Similar listings buyers compare against this one.

Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks

Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0

Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection

Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure

JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0

Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio

Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026

Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms