BlogLLM Benchmark Scores as Published, Side by Side

Qwen3-VL-235B-A22B: Benchmark Scores as Published

Every result the catalog holds for Qwen3-VL-235B-A22B: 46 tests in 3 areas, from 2 pages. All were printed by Alibaba itself. Each number is as printed; nothing is ranked or rescored.

Developer AlibabaReleased 2025-09-23Data version 2026103001, as of 2026-10-30

Computer Use

TestValuePrinted bySetup as printed
OSWorld-Verified38.1%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
ScreenSpot-Pro62%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
AndroidWorld63.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-

Multimodal Understanding

TestValuePrinted bySetup as printed
CharXiv Reasoning (no tools)66.1%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMMU-Pro69.3%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMMU (validation)80.6%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MathVision74.6%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MathVista (testmini)85.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
AI2D (test)89.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
LVBench63.6%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
OmniDocBench 1.584.5%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
Video-MMMU80%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
OCRBench87.5%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
HallusionBench66.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMStar78.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
BabyVision (no tools)22.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
ERQA52.5%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
SimpleVQA61.3%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
Video-MME (with subtitles)83.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
RealWorldQA81.3%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMVU71.1%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
V* Bench (tool setting not stated)85.9%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
CC-OCR81.5%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMBench-EN dev v1.189.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
CountBench93.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
DynaMath82.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
ZeroBench (subquestions)28.4%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
EmbSpatial-Bench84.3%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MLVU83.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MMLongBench-Doc56.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
Video-MME (without subtitles)79%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
ZeroBench (metric not stated)4%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
MVBench75.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
RefSpatial-Bench69.9%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
LingoQA66.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
RefCOCO (average)91.1%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
VLMs are Blind79.5%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15-
ODinW-1343.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
We-Math74.8%Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15-
Hypersim (metric not stated)11Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
nuScenes (metric not stated)13.9Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
SUN RGB-D (metric not stated)34.9Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
TIR-Bench (tool setting not stated)29.8%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15-

Healthcare

TestValuePrinted bySetup as printed
MedXpertQA MM47.6%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
PMC-VQA41.2%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-
SLAKE54.7%Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page-

More from Alibaba

Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.