Which is better, Qwen3.5 2B or Qwen3.5 4B?

Qwen3.5 2B vs Qwen3.5 4B

43 SHARED BENCHMARKS

Across 43 shared benchmarks, Qwen3.5 2B scores higher on 1 and Qwen3.5 4B on 41, with 1 level. The widest gap is Terminal-Bench 2.1, where Qwen3.5 4B scores 25.8 against 3.

ALIBABAVSALIBABA43 SHARED1–41 HEAD-TO-HEAD

BenchmarkQwen3.5 2BMARGINQwen3.5 4B

AA Agentic Index0.836.3

AA Intelligence6.920

AA-LCR25.663

AA-Omniscience-60.2-58.4

Artificial Analysis Coding Index2.922.6

bfcl_v443.654

C-Eval73.285.1

critpt00

gdpval08.3

GPQA Diamond51.677.1

HLE59.9

IFBench41.359.2

ifeval78.689.8

include55.471

longbench_v238.750

MathVista48.763.6

MM-IFEval55.463.1

MMBench73.187.1

MME76.279.5

mmlu_prox52.371.5

mmlu_redux79.688.8

MMLU-Pro66.579.1

mmmlu63.176.1

MMMU (val) (Pass@1)44.150.3

MMMU-Pro4365.4

MMStar55.175.3

multichallenge33.749

Multilingual MMBench69.977

Multilingual MMMB75.982

OCRBench86.186.9

OmniScience Accuracy8.115.1

OmniScience Non-Hallucination25.813.4

RealWorldQA65.176.3

scicode7.218.3

SEED-Bench (image)75.876.1

SimpleVQA35.247.8

supergpqa37.552.9

t2-bench48.879.9

TauBench V3 - Banking2.56.8

Terminal-Bench 2.1325.8

Terminal-Bench Hard3.818.2

τ²-Bench Telecom (AA run)81.692.1

τ³-Bench6.28.2

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.

Compare Qwen3.5 2B withALL PAIRINGS →

AlibabaQwen3.5 35B A3B Qwen3 VL 4B (Reasoning)Qwen3.5 397B A17B Qwen2.5 Instruct 72B Qwen3.5 27B Qwen3.6 Plus

Across the fieldGemini 3.6 Flash Nemotron 3 Super 120B A12B GPT-5.5 Claude Opus 4.8 GPT-4o GPT-5.6 Sol Gemini 3.1 Pro GPT-5.4

Compare Qwen3.5 4B withALL PAIRINGS →

AlibabaQwen3.5 35B A3B Qwen3 VL 4B (Reasoning)Qwen3.5 397B A17B Qwen2.5 Instruct 72B Qwen3.5 27B Qwen3.6 Plus

Across the fieldQwen3.6 27B Qwen3.7 Max Qwen3.8 Max Preview Qwen3.7 Plus Preview Claude Opus 4 Gemini 3.6 Flash Nemotron 3 Super 120B A12B GPT-5.5