AI Radar

DeepSeek-R1-Zero

Available · Language, Reasoning

DeepSeek-R1-Zero is one of DeepSeek's first-generation reasoning models, trained with large-scale reinforcement learning directly on DeepSeek-V3-Base without supervised fine-tuning first, and released as open weights under the MIT License alongside DeepSeek-R1. DeepSeek names endless repetition, poor readability and language mixing as its weaknesses. [1] [2]Introduction [3] Primary source

Timeline of DeepSeek-R1-Zero →

Claims and evidence

  • Released No DeepSeek page names DeepSeek-R1-Zero together with a date. It was open-sourced with DeepSeek-R1: the commit Release DeepSeek-R1 of 2025-01-20 adds the README that announces both models, and the R1 announcement of that day says the open-source model is out. Primary source[4]commit date [3]1. Introduction; 3. Model Downloads [5]news sidebar: DeepSeek-R1 Release 2025/01/20
  • Status AvailableWeights are still public on Hugging Face. Primary source[1]
  • Derived from (other) DeepSeek-V3Trained from the base variant DeepSeek-V3-Base with large-scale reinforcement learning (GRPO), without supervised fine-tuning as a preliminary step. Primary source[1]2. Model Summary; 3. Model Downloads [2]Introduction
  • Change · Reasoning Reasoning is learned through large-scale reinforcement learning (GRPO) applied directly to DeepSeek-V3-Base, with no supervised fine-tuning stage first.Compared with DeepSeek-V3 Primary source[2] [1]
  • Open weights YesMIT License. Primary source[1]1. Introduction; 7. License [3]7. License
  • Context window 128K tokens Primary source[1]3. Model Downloads [3]3. Model Downloads
  • Parameters 671B total, 37B activatedMixture-of-experts. Primary source[1]3. Model Downloads [3]3. Model Downloads
  • Access Open-weights download Primary source[1]3. Model Downloads

Lineage

Predecessors

No known predecessor.

Successors

No known successor.

Based on

Variants and derived

None recorded.

Siblings

None recorded.

All ancestors

All descendants

None.

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from DeepSeek on AI Radar →