DeepSeek R1 Distill Qwen 14B

Overview

DeepSeek's R1 reasoning distilled into Qwen 14B under MIT. AIME24 69.7 and MATH-500 93.9 — beats o1-mini on most reasoning benchmarks.

When to pick this model

• Math, coding, and STEM reasoning on a single 24GB GPU
• Local alternative to o1-mini-class APIs
• Workloads needing MIT-licensed reasoning
• Agentic planners that benefit from explicit CoT

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	9 GB
Q5_K_M	11 GB
Q8_0	16 GB
FP16 (no quantization)	28 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Published benchmark scores

Benchmark	Score
AIME 2024	69.7
MATH-500	93.9
GPQA	59.1

Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.

Strengths

• AIME24 69.7 and MATH-500 93.9
• Beats o1-mini on multiple reasoning benchmarks
• MIT license — no usage restrictions
• 131k context

Limitations

• Verbose CoT inflates token costs
• Slower than non-reasoning 14B for simple queries
• No vision or tool-use specialization

Architecture & training

Architecture: Dense Qwen 2.5 14B · SFT on R1 traces

Training: Distillation of R1 671B.

Quick start

ollama run deepseek-r1:14b

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.