whatmodel.sh / DeepSeek R1 Distill Qwen 14B

DeepSeek R1 Distill Qwen 14B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

DeepSeek R1 Distill Qwen 14B is an open-weights language model from DeepSeek, released January 2025. Its published size is 14.8B parameters and a 131k-token context window. Licence: mit.

55.7±1.1#252 of 719 LLMsfrom 18 benchmark results across 8 sources

$ how it compares

The LLMs ranked either side of DeepSeek R1 Distill Qwen 14B, out of 828 with a score, then more from DeepSeek.

#modelscorewhereRAMreleased
301DeepSeek-V2.5 (1210) DeepSeek56±1local~82GB2024-12
302Qwen3 8B Alibaba56±1api $0.31/M +local~6.1GB2025-04
303DeepSeek R1 Distill Qwen 14B DeepSeek56±1local~10GB2025-01
304Ministral 3 14B Mistral56±1api $0.20/M +local~9.1GB2025-12
305Nemotron Nano 9B v2 NVIDIA56±1api $0.10/M +local~7.6GB2025-08
more from DeepSeek
21DeepSeek V4.1 Flash DeepSeek78±1api $0.26/M +local~537GB2026-09
31DeepSeek V4 Pro (0813) DeepSeek76±1api $1.63/M +local~1010GB2026-08
35DeepSeek V3.2 Speciale DeepSeek76±3comparison only

see all LLMs · find the ones that fit your machine

$ what it's good at

coding▁▁▁▃▆██▇▆▄▃▁ #237 of 458 · 18% of its score · 5 tests

Beats Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B), GLM-4.7-Flash here, though they rank higher overall. Trails QwQ-32B-Preview, InternLM2.5 20B Chat, Gemma 2 27B IT here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LiveCodeBench (version unspecified) lab53.1% ✎▄····▄▄··▄█▄ beats 3/6medium · avg 41%▇▇▇▇ 9.2%at its level
CodeReviewQA: change localisation (hard), invariant accuracy53.2% ✎█▂▁▁▂▂▂▃▅▅▅▇ beats 37/47medium · avg 36%▇ 3.6%sharp · at its level · noisy
CodeReviewQA: automated code refinement, exact match rate31.3% ✎▁▁▁▁▁▁▂▂▃█▅▄ beats 25/47hard · avg 33%▇ 2.4%at its level
CodeReviewQA: solution identification (hard), invariant accuracy66.8% ✎▃··▁▂▂▃▂█▃▄▂ beats 36/47medium · avg 54%▇ 2.3%sharp · at its level · noisy
CodeReviewQA: change type recognition, invariant accuracy72.7% ✎▁▁▁······▁▃█ beats 23/47easy · avg 67%▇ 0.0%blunt · at its level · noisy
math▁▂▃▅▆▅▄█▄▁▁▁ #172 of 398 · 40% of its score · 5 tests

Beats MiniMax-M2.1, Mistral Medium 3.5 (128B), GPT-4.1 here, though they rank higher overall. Trails OpenThinker3 7B, Olmo 3 7B Think, OpenThinker2 32B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MATH level 587.1% ✎▂▂▅▆▄█▆▃▅▂▃▄ beats 73/92hard · avg 32%▇▇▇▇▇▇▇▇▇▇ 24%sharp · at its level
OTIS mock AIME 2024–202550.6% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 76/193very hard · avg 8.5%▇▇▇ 7.5%sharp · noisy
AIME 2024 lab69.7% ✎▃·▃·▃▃···█·▃ beats 4/7medium · avg 50%▇ 3.4%sharp · at its level · noisy
MATH-500 lab93.9% ✎▄·▄▄·▄▄█···▄ beats 4/7very easy · avg 91%▇ 2.8%maxed out
THINK-Bench final-answer accuracy (reasoning models)70.2% ✎▄█·▄····▄▄·█ beats 2/7easy · avg 76%▇ 2.1%typical
other tests7 not in a task · 42% of its score
testresultvs the others in its rundifficultycountswhy
GPQA Diamond +144.7% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 49/208hard · avg 40%▇▇▇▇▇▇▇▇▇▇▇ 29%sharp · at its level · several runs
BRIDGE clinical text, zero-shot average (87 tasks)34.3% ✎▁▁▁▁▁▂▂▃▄▆█▃ beats 43/97hard · avg 32%▇▇ 5.4%blunt · at its level
SurgiQ surgical MCQ (zero-shot, log-probability scoring)55.9% ✎▂···▂▂·▃▂▃██ beats 7/17medium · avg 53%▇ 2.5%at its level
Open PT-LLM Leaderboard average (9 Portuguese tasks)68.1% ✎▃▁▁▁▁▁▁▁▂█▅▁ beats 147/276medium · avg 73%▇ 1.9%at its level · noisy
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy72.6% ✎▃·▁▁▂▁▂▅█▇▄▁ beats 57/100medium · avg 68%▇ 1.5%at its level · noisy
Chess Puzzles (Epoch)1.0% ✎█▁▁▁▁▂▁▄▂▂▁▁ beats 26/135very hard · avg 5.9%▇ 0.9%sharp · too hard for it · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy53.8% ✎▁▁▁▁▁▁▁▃▄█▄▃ beats 23/146easy · avg 83%▇ 0.8%noisy

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run DeepSeek R1 Distill Qwen 14B · needs ~9 GB · from huggingface.co/lmstudio-community
    chat in the terminal
    llama cli -hf lmstudio-community/DeepSeek-R1-Distill-Qwen-14B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf lmstudio-community/DeepSeek-R1-Distill-Qwen-14B-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run DeepSeek R1 Distill Qwen 14B · needs ~9 GB · from huggingface.co/lmstudio-community

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/lmstudio-community/DeepSeek-R1-Distill-Qwen-14B-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run DeepSeek R1 Distill Qwen 14B · needs ~9.38 GB · from ollama.com
    chat in the terminal
    ollama run deepseek-r1:14b

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run DeepSeek R1 Distill Qwen 14B · needs ~8.3 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/DeepSeek-R1-Distill-Qwen-14B-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/DeepSeek-R1-Distill-Qwen-14B-4bit

Weights and documentation: huggingface.co/deepseek-ai ↗

$ how the score was worked out · Epoch AI — internal evaluatio… 50%, deepseek-ai/DeepSeek-R1 model… 27%, CodeReviewQA: The Code Review… 8% and 5 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Epoch AI — internal evaluations4▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 50%CC BY 4.0
deepseek-ai/DeepSeek-R1 model card · lab-reported4▇▇▇▇▇▇▇▇ 27%mit
CodeReviewQA: The Code Review Comprehension Assessment for Large Language Models (arXiv:2503.16167v2)4▇▇▇ 8%CC-BY-4.0
BRIDGE Medical Leaderboard (Harvard Medical School / Brigham and Women's YLab)1▇▇ 5%MIT
SurgiQ: A Large-Scale Multi-Domain Benchmark for Evaluating Surgical Understanding in Large Language Models (arXiv:2606.08071v1)1▇ 2%CC BY 4.0
EuroEval leaderboards (results collection), English generative leaderboard2▇ 2%MIT
THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models (arXiv:2505.22113v1)1▇ 2%CC-BY-4.0
Open Portuguese LLM Leaderboard (archived)1▇ 2%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 55.7
$ quantization evidence · 27 measurements

How much of the full-precision result a quantized version kept, measured within one study. Rows from the same study share a table and aren't independent.

precisiontestfull → quantizedkept
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)AIME 2024 pass@10.667 → 0.681102.2%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)IFEval 0-shot instruction-level strict acc (OpenLLM v2 config, chat template)0.43 → 0.437101.5%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)GPQA Diamond pass@10.594 → 0.601101.3%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)Reasoning average of 3 tasks (AIME 2024 pass@1, MATH-500 pass@1, GPQA Diamond pass@1)0.736 → 0.743101.0%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)OpenLLM v2 average of 6 tasks (IFEval, BBH, MATH-Hard, GPQA, MuSR, MMLU-Pro; lm-evaluation-harness `leaderboard`, chat template, vLLM; values exactly as published, per-task normalisation not stated)0.342 → 0.344100.5%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)MMLU-Pro 5-shot (OpenLLM v2 config; value as published, normalisation not stated)0.349 → 0.35100.5%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)GSM8K 5-shot strict-match (lm-evaluation-harness, vLLM)0.87 → 0.874100.4%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)MMLU 5-shot acc (lm-evaluation-harness, vLLM)0.745 → 0.746100.2%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)MATH-500 pass@10.947 → 0.946100.0%source ↗
FP8 W8A8 dynamic (FP8 weights symmetric per-channel, FP8 activations dynamic per-token; llm-compressor FP8_DYNAMIC)OpenLLM v1 average of 6 tasks (ARC-Challenge 25-shot, GSM8K 5-shot, HellaSwag 10-shot, MMLU 5-shot, TruthfulQA MC2 0-shot, Winogrande 5-shot; lm-evaluation-harness `openllm`, vLLM)0.71 → 0.70899.7%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)MMLU-Pro 5-shot (OpenLLM v2 config; value as published, normalisation not stated)0.349 → 0.351100.7%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)GSM8K 5-shot strict-match (lm-evaluation-harness, vLLM)0.87 → 0.873100.3%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)MATH-500 pass@10.947 → 0.95100.3%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)OpenLLM v1 average of 6 tasks (ARC-Challenge 25-shot, GSM8K 5-shot, HellaSwag 10-shot, MMLU 5-shot, TruthfulQA MC2 0-shot, Winogrande 5-shot; lm-evaluation-harness `openllm`, vLLM)0.71 → 0.70999.8%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)AIME 2024 pass@10.667 → 0.6699.1%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)Reasoning average of 3 tasks (AIME 2024 pass@1, MATH-500 pass@1, GPQA Diamond pass@1)0.736 → 0.72899.0%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)MMLU 5-shot acc (lm-evaluation-harness, vLLM)0.745 → 0.73398.5%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)GPQA Diamond pass@10.594 → 0.57596.8%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)OpenLLM v2 average of 6 tasks (IFEval, BBH, MATH-Hard, GPQA, MuSR, MMLU-Pro; lm-evaluation-harness `leaderboard`, chat template, vLLM; values exactly as published, per-task normalisation not stated)0.342 → 0.32494.7%source ↗
INT4 W4A16 (weights only, symmetric per-group, group size 128; card states GPTQ via llm-compressor)IFEval 0-shot instruction-level strict acc (OpenLLM v2 config, chat template)0.43 → 0.34981.1%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)GSM8K 5-shot strict-match (lm-evaluation-harness, vLLM)0.87 → 0.878100.9%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)OpenLLM v1 average of 6 tasks (ARC-Challenge 25-shot, GSM8K 5-shot, HellaSwag 10-shot, MMLU 5-shot, TruthfulQA MC2 0-shot, Winogrande 5-shot; lm-evaluation-harness `openllm`, vLLM)0.71 → 0.711100.1%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)MATH-500 pass@10.947 → 0.947100.0%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)MMLU 5-shot acc (lm-evaluation-harness, vLLM)0.745 → 0.74399.7%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)AIME 2024 pass@10.667 → 0.66399.5%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)Reasoning average of 3 tasks (AIME 2024 pass@1, MATH-500 pass@1, GPQA Diamond pass@1)0.736 → 0.73199.4%source ↗
INT8 W8A8 (weights symmetric per-channel, activations symmetric per-token; SmoothQuant + GPTQ via llm-compressor)GPQA Diamond pass@10.594 → 0.58398.3%source ↗

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json