HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks
Home/Orgs/Qwen/Qwen/Qwen3-4B-Instruct-2507

Qwen/Qwen3-4B-Instruct-2507

family qwen · 22 variants · unknown

provider
Qwen
released
—
huggingface
https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507
best scores
MMLU-Pro: 62.8MMLU: 80.2GPQA: 60.26AIME 2025: 48.75LiveCodeBench: 31.74IFEval: 74.5Chatbot Arena: 15.9BFCL (function calling): 53.0AIME 2024: 64.79MATH: 94.66CodeContests: 42.08Codeforces: 33.69HumanEval: 77.56

Best scores per benchmark

benchmarkcategorybest score
MATHmath94.66
MMLUknowledge80.2
HumanEvalcoding77.56
IFEvalinstruction-following74.5
AIME 2024math64.79
MMLU-Proknowledge62.8
GPQAreasoning60.26
BFCL (function calling)agentic53.0
AIME 2025math48.75
CodeContestscoding42.08
Codeforcescoding33.69
LiveCodeBenchcoding31.74
Chatbot Arenachat15.9

Variants (22)

variantkindquantbitsvRAMdlorgyear
Qwen3-4B-Instruct-2507originalbf1616.03,506,956Qwen2025
Qwen3-4B-Instruct-2507-FP8fp8fp88.01,037,749Qwen2025
Qwen3-4B-Instruct-2507-GGUFggufgguf16.078,387unsloth2025
Qwen3-4B-Instruct-2507-unsloth-bnb-4bitbnb-4bit4bit76,738unsloth2025
Qwen3-4B-Instruct-2507-MLX-4bitmlxmlx52,095lmstudio-community2025
Qwen3-4B-Instruct-2507-MLX-8bitmlxmlx47,737lmstudio-community2025
Qwen3-4B-Instruct-2507-MLX-5bitmlxmlx47,154lmstudio-community2025
Qwen3-4B-Instruct-2507-MLX-6bitmlxmlx47,053lmstudio-community2025
Qwen3-4B-Instruct-2507originalbf1616.038,537unsloth2025
Qwen3-4B-Instruct-2507-GGUFggufgguf3.021,568lmstudio-community2025
Qwen_Qwen3-4B-Instruct-2507-GGUFggufgguf2.012,842bartowski2025
Qwen3-4B-Instruct-2507-bnb-4bitbnb-4bit4bit7,561unsloth2025
Qwen3-4B-Instruct-2507-Q8_0-GGUFggufgguf8.01,618ggml-org2025
Privasis-Cleaner-4Boriginalbf1616.0935nvidia2026
AesCoder-4B-GGUFggufgguf16.0516Mungert2025
Qwen3-4B-Instruct-2507-GGUFggufgguf16.0347Mungert2025
WEBGEN-4B-Preview-GGUFggufgguf16.0121Mungert2025
Nemotron-Research-GooseReason-4B-Instructoriginalbf1616.0103nvidia2026
intent-aware-lfqa-qwen3-4b-intent-explicitoriginalbf1616.064allenai2026
intent-aware-lfqa-qwen3-4b-intent-implicitoriginalbf1616.063allenai2026
intent-aware-lfqa-qwen3-4b-baselineoriginalbf1616.062allenai2026
intent-aware-lfqa-qwen3-4b-multivieworiginalbf1616.056allenai2026
Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.