# DeepSeek-R1-Distill-Qwen-7B

> DeepSeek-R1 reasoning distilled into a 7B Qwen model you can run locally.

- Maker: DeepSeek
- Line: DeepSeek R1 Distill
- Canonical page: https://ai-tldr.dev/models/deepseek-r1-distill-qwen-7b/
- Updated: 2026-06-18

## In short

DeepSeek-R1-Distill-Qwen-7B is one of six distilled models DeepSeek released on 20 January 2025 alongside the full DeepSeek-R1. It takes Alibaba's Qwen2.5-Math-7B as the base and fine-tunes it on roughly 800,000 reasoning traces generated by the 671B-parameter DeepSeek-R1. You can use it through Hugging Face (weights), OpenRouter and Fireworks AI.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 131,072 tokens (max_position_embeddings in config.json) (max output 32,768 tokens recommended for the R1-distill series (long chain-of-thought generations)) | [1] |
| Licence | MIT (repository); derived from Qwen2.5-Math-7B, originally Apache 2.0. Commercial use, modification and further distillation are permitted. · open weights | [1] |
| Released | 20 January 2025 | [2] |
| Parameters | 7.6B (dense) | [1] |
| Knowledge cutoff | Not officially published by DeepSeek; inherits the underlying Qwen2.5 / DeepSeek-R1 training data. | [1] |

## Sources

1. [Hugging Face model card](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B)
2. [DeepSeek-R1 GitHub (benchmarks & license)](https://github.com/deepseek-ai/DeepSeek-R1)

## Overview

DeepSeek-R1-Distill-Qwen-7B is one of six distilled models DeepSeek released on 20 January 2025 alongside the full DeepSeek-R1. It takes Alibaba's Qwen2.5-Math-7B as the base and fine-tunes it on roughly 800,000 reasoning traces generated by the 671B-parameter DeepSeek-R1. The result is a small, dense 7B model that inherits much of R1's step-by-step "thinking" behaviour — it writes out a long chain of thought before giving a final answer — while being light enough to run on a single consumer GPU.

Because it is built on the math-specialised Qwen2.5-Math-7B backbone, the model is strongest at mathematics and competition-style reasoning. DeepSeek reports 92.8% on MATH-500 and 55.5% on AIME 2024 for this 7B distill, results that the team noted surpass the much larger QwQ-32B-Preview on those tasks. It keeps the standard Qwen2 architecture (28 layers, grouped-query attention, 152k vocabulary) and a 131,072-token context window.

The weights are openly published on Hugging Face under an MIT repository license, and the model derives from Qwen2.5 (Apache 2.0), so commercial use, modification and further distillation are all permitted. DeepSeek advises running it with no system prompt — all instructions go in the user message — and a temperature around 0.6 to keep the reasoning stable. There is no first-party paid API; the model is meant to be self-hosted or run through third-party inference providers.

## Benchmarks

- AIME 2024 (pass@1): 55.5% ([source](https://github.com/deepseek-ai/DeepSeek-R1))
- MATH-500 (pass@1): 92.8% ([source](https://github.com/deepseek-ai/DeepSeek-R1))
- GPQA Diamond (pass@1): 49.1% ([source](https://github.com/deepseek-ai/DeepSeek-R1))
- LiveCodeBench (pass@1): 37.6% ([source](https://github.com/deepseek-ai/DeepSeek-R1))

## Strengths

- Strong math and competition reasoning for its size — 92.8% on MATH-500 and 55.5% on AIME 2024, beating some 32B models
- Small enough to run locally: a 7.6B dense model fits comfortably on a single consumer GPU (roughly 6-16 GB VRAM depending on quantization)
- Open weights under a permissive MIT/Apache-2.0 chain — free to self-host, fine-tune, and use commercially
- Inherits DeepSeek-R1's explicit chain-of-thought style, useful for transparent step-by-step reasoning
- Large 131K-token context window for a model this size
- Widely supported across inference stacks (vLLM, SGLang, llama.cpp/Ollama, LM Studio) and many third-party API hosts

## Best for

- Local and offline math, logic, and reasoning assistants on a single GPU
- Cost-sensitive reasoning workloads where a 7B model is enough
- Education: showing worked, step-by-step solutions to math and competition problems
- On-device coding help and algorithmic problem-solving (e.g. LeetCode-style tasks)
- A fine-tuning base for domain-specific reasoning models
- Research and experimentation with distilled chain-of-thought reasoning

## How to access

- [Hugging Face (weights)](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B) — model id `deepseek-ai/DeepSeek-R1-Distill-Qwen-7B`
- [OpenRouter](https://openrouter.ai/deepseek/deepseek-r1-distill-qwen-7b) — model id `deepseek/deepseek-r1-distill-qwen-7b`
- [Fireworks AI](https://fireworks.ai/models/fireworks/deepseek-r1-distill-qwen-7b) — model id `accounts/fireworks/models/deepseek-r1-distill-qwen-7b`
- [NVIDIA NIM](https://build.nvidia.com/deepseek-ai/deepseek-r1-distill-qwen-7b) — model id `deepseek-ai/deepseek-r1-distill-qwen-7b`

## DeepSeek R1 Distill versions

- [DeepSeek-R1-0528-Qwen3-8B](https://ai-tldr.dev/models/deepseek-r1-0528-qwen3-8b/) (29 May 2025)
- [DeepSeek-R1-Distill-Llama-70B](https://ai-tldr.dev/models/deepseek-r1-distill-llama-70b/) (20 January 2025)
- [DeepSeek-R1-Distill-Qwen-32B](https://ai-tldr.dev/models/deepseek-r1-distill-qwen-32b/) (20 January 2025)
- [DeepSeek-R1-Distill-Qwen-14B](https://ai-tldr.dev/models/deepseek-r1-distill-qwen-14b/) (20 January 2025)
- [DeepSeek-R1-Distill-Llama-8B](https://ai-tldr.dev/models/deepseek-r1-distill-llama-8b/) (20 January 2025)
- DeepSeek-R1-Distill-Qwen-7B (20 January 2025) — this page
- [DeepSeek-R1-Distill-Qwen-1.5B](https://ai-tldr.dev/models/deepseek-r1-distill-qwen-1-5b/) (20 January 2025)

## FAQ

### What is DeepSeek-R1-Distill-Qwen-7B based on?

It is a fine-tune of Alibaba's Qwen2.5-Math-7B. DeepSeek distilled the reasoning behaviour of its full 671B DeepSeek-R1 model into it using about 800,000 supervised reasoning samples generated by R1. It keeps the standard Qwen2 dense architecture (28 layers, grouped-query attention).

### How good is it at math and coding?

Very strong for a 7B model on math: DeepSeek reports 92.8% on MATH-500 and 55.5% on AIME 2024, which it notes beats the larger QwQ-32B-Preview on those tasks. On coding it scores 37.6% pass@1 on LiveCodeBench, and 49.1% on GPQA Diamond for science reasoning.

### What context window and license does it have?

The config sets a 131,072-token context window. The weights ship under an MIT repository license and the model derives from Qwen2.5 (Apache 2.0), so commercial use, modification and further distillation are all allowed.

### How much does it cost to use?

DeepSeek does not sell a first-party API for this model — it is open weights, so you can download and self-host it for free. If you do not want to run it yourself, third-party hosts such as OpenRouter, Fireworks AI and NVIDIA NIM serve it; their per-token prices vary, so check the provider for the current rate.

## Links

- [Hugging Face model card](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B)
- [DeepSeek-R1 GitHub (benchmarks & license)](https://github.com/deepseek-ai/DeepSeek-R1)
- [DeepSeek-R1 paper (arXiv 2501.12948)](https://arxiv.org/html/2501.12948v1)
- [OpenRouter — R1 Distill Qwen 7B](https://openrouter.ai/deepseek/deepseek-r1-distill-qwen-7b)
- [Fireworks AI — R1 Distill Qwen 7B](https://fireworks.ai/models/fireworks/deepseek-r1-distill-qwen-7b)
- [NVIDIA NIM — R1 Distill Qwen 7B](https://build.nvidia.com/deepseek-ai/deepseek-r1-distill-qwen-7b)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/deepseek-r1-distill-qwen-7b/