DeepSeek-R1-Distill-Qwen-14B | Epoch AI

ECI

N/A

DeepSeek

Jan. 20, 2025

Open weights

DeepSeek-R1-Distill-Qwen-14B is an open-weights model from DeepSeek, released January 20, 2025.

DeepSeek-R1-Distill-Qwen-14B vs select alternatives

Domain	Benchmark	DeepSeek-R1-Distill-Qwen-14B DeepSeek Jan. 20, 2025 Open weights	o1 OpenAI Dec. 17, 2024 Closed weights	Gemini 2.0 Flash (Dec 2024) Google DeepMind Dec. 11, 2024 Closed weights	Claude 3.5 Sonnet (October 2024) Anthropic Oct. 22, 2024 Closed weights
Mathematics	MATH Level 5Best score 98%	87%	95%	n/a	57%
	OTIS Mock AIME 2024-2025Best score 100%	51%	73%	n/a	8%
Science	GPQA DiamondBest score 96%	45%	77%	n/a	55%

Model metadata

General
Organization	DeepSeek
Release date	Jan. 20, 2025
Model accessibility	Open weights
Country	China
Knowledge cutoff	Unknown
Provider notes	DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

Pricing
Input price	$0.15 / 1M tokens
Output price	$0.15 / 1M tokens

Training
Parameters	1.5e10
Training compute	Unknown
Training hardware	Unknown
Training time	Unknown
Training dataset size	2.4e10 datapoints