DeepSeek's open-weights reasoning LLM (671B MoE, 37B active) trained with large-scale reinforcement learning.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
DeepSeek-R1 is a first-generation reasoning model from DeepSeek, built on DeepSeek-V3-Base as a mixture-of-experts model with 671B total and 37B activated parameters and a 128K context length. It was released alongside DeepSeek-R1-Zero, which was trained with large-scale reinforcement learning without a supervised fine-tuning step. R1 adds cold-start data and a pipeline of two RL and two SFT stages to address R1-Zero's repetition, poor readability and language mixing.
According to the vendor, R1 performs comparably to OpenAI-o1 on math, code and reasoning benchmarks. Its accompanying paper described training reasoning with reinforcement learning using GRPO (group relative policy optimization), which scores sampled answers against the group average. DeepSeek also released six dense models distilled from R1 (1.5B, 7B, 8B, 14B, 32B, 70B) based on Qwen2.5 and Llama 3 models. The distilled models inherit the licenses of their base models in part.
The weights are published on Hugging Face under the MIT license, so the model can be self-hosted, and it is also available through DeepSeek's chat app and API. The README recommends reading its usage recommendations before running locally. The model is text-only; the full-size model needs substantial GPU resources to run. DeepSeek's website now also lists newer models such as DeepSeek-V4.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Its paper showed publicly how to train reasoning with reinforcement learning using GRPO (group relative policy optimization). GRPO samples many answers per problem, scores them against the group average (advantage), and reinforces better-than-average answers.
Unitree's lightest and cheapest humanoid, starting at $4,900, shown in a demo doing acrobatic breakdancing moves, flips and stable landings.
Closest neighbours by meaning, not just by category.
Model · DeepSeek
Chinese AI lab behind the open-weight DeepSeek-V4 and DeepSeek-R1 models, with a chat app and low-cost API.
Model · OpenAI
OpenAI's first reasoning model series, trained with reinforcement learning to think before answering; now deprecated in the API.
Model · OpenAI
OpenAI's open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, released under Apache 2.0.
Model
Open 7.39B dense model for math reasoning and agentic search, with a 256K context and the full training pipeline released.
Model · DeepSeek
Open-weights multimodal DeepSeek model in the V4 family, offered via API, chat and downloadable weights.
Model · DeepSeek
Open-weights Flash-tier model in DeepSeek's V4 family, offered via chat and a low-cost API.
