Which is better, Qwen3.5 2B or Qwen3.5 35B A3B?

Qwen3.5 2B vs Qwen3.5 35B A3B

37 SHARED BENCHMARKS

Across 37 shared benchmarks, Qwen3.5 2B scores higher on 1 and Qwen3.5 35B A3B on 36. The widest gap is HLE, where Qwen3.5 35B A3B scores 47.4 against 5.

ALIBABAVSALIBABA37 SHARED1–36 HEAD-TO-HEAD

BenchmarkQwen3.5 2BMARGINQwen3.5 35B A3B

AA Agentic Index0.844.1

AA Intelligence6.919.3

AA-LCR25.672

AA-Omniscience-60.2-48.1

Artificial Analysis Coding Index2.937

bfcl_v443.667.3

C-Eval73.290.2

critpt00.9

gdpval020.3

GPQA Diamond51.684.5

HallusionBench65.567.9

HLE547.4

IFBench41.372.5

ifeval78.691.9

include55.479.7

longbench_v238.759

MathVista48.786.2

mmlu_prox52.381

mmlu_redux79.693.3

MMLU-Pro66.585.3

mmmlu63.185.2

MMMU-Pro4375.1

MMStar55.181.9

multichallenge33.760

OCRBench86.191

OmniScience Accuracy8.120.1

OmniScience Non-Hallucination25.814.6

RealWorldQA65.184.1

scicode7.237.7

SimpleVQA35.258.3

supergpqa37.563.4

t2-bench48.881.2

TauBench V3 - Banking2.54.9

Terminal-Bench 2.1340.8

Terminal-Bench Hard3.826.5

τ²-Bench Telecom (AA run)81.689.2

τ³-Bench6.268.9

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

Across the fieldDeepSeek-V4-Pro DeepSeek-V3 Kimi K2.6 Kimi K2.5 o3 GPT-5.1 Claude Opus 5 Gemma 4 26B A4B

Compare Qwen3.5 35B A3B withALL PAIRINGS →

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

Across the fieldQwen3.6 27B Qwen3.7 Max Qwen3.8 Max Preview Qwen3.7 Plus Preview Claude Opus 4 Qwen3.5 4B Gemini 3.6 Flash Nemotron 3 Super 120B A12B