HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks
Home/Orgs/Qwen/Qwen/Qwen2.5-14B

Qwen/Qwen2.5-14B

family qwen · 16 variants · unknown

provider
Qwen
released
—
huggingface
https://huggingface.co/Qwen/Qwen2.5-14B
best scores
GPQA: 31.6MMLU-Pro: 47.5Humanity's Last Exam: 11.8LiveCodeBench: 28.6SciCode: 100.0AIME 2024: 83.3AIME 2025: 70.0

Best scores per benchmark

benchmarkcategorybest score
SciCodecoding100.0
AIME 2024math83.3
AIME 2025math70.0
MMLU-Proknowledge47.5
GPQAreasoning31.6
LiveCodeBenchcoding28.6
Humanity's Last Examknowledge11.8

Variants (16)

variantkindquantbitsvRAMdlorgyear
Qwen2.5-14B-Instructoriginalbf1616.03,374,179Qwen2024
Qwen2.5-14B-bnb-4bitbnb-4bit4bit218,948unsloth2025
Qwen2.5-14Boriginalbf1616.089,920Qwen2024
Qwen2.5-Coder-14Boriginalbf1616.066,050Qwen2024
Qwen2.5-14B-Instruct-1Moriginalbf1616.09,864Qwen2025
Qwen2.5-14B-unsloth-bnb-4bitbnb-4bit4bit2,528unsloth2025
OpenReasoning-Nemotron-14Boriginalbf1616.01,843nvidia2025
Qwen2.5-14Boriginalbf1616.01,357unsloth2025
Qwen2.5-14B-i1-GGUFggufgguf1.01,033mradermacher2024
Qwen2.5-14B-Instruct-1M-GGUFggufgguf16.0246Mungert2025
Qwen2.5-14B-Instruct-GGUFggufgguf16.0188Mungert2025
OpenMath-Nemotron-14Boriginalbf1616.0150nvidia2025
OpenMath-Nemotron-14B-Kaggleoriginalbf1616.077nvidia2025
OpenMath-Nemotron-14B-Kaggle-GGUFggufgguf16.037Mungert2025
ASearcher-Local-14Boriginalbf1616.024inclusionAI2025
ASearcher-Web-14Boriginalbf1616.022inclusionAI2025
Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.