Qwen3.5 122B A10B

Name: Qwen3.5 122B A10B
Author: Alibaba

Alibaba

Qwen3.5 122B A10B is a Mixture-of-Experts model from the Qwen3.5 series. 122 billion parameters with 10 billion active during inference.

Key Specifications

Parameters

122.0B

Context

Release Date

March 1, 2026

Average Score

70.9%

API Documentation Repository Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

March 1, 2026

Last Update

March 21, 2026

Today

October 5, 2026

Technical Specifications

Parameters

122.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

Self-reported by the model provider • Self-reported

72.0%

Reasoning

Logical reasoning and analysis

GPQA

Self-reported by the model provider • Self-reported

86.6%

Multimodal

Working with images and visual data

AI2D

TEST • Self-reported

93.0%

MMMU

Self-reported by the model provider • Self-reported

83.9%

Other Tests

Specialized benchmarks

CountBench

• Self-reported

97.0%

VLMsAreBlind

• Self-reported

97.0%

MMLU-Redux

• Self-reported

94.0%

IFEval

• Self-reported

93.0%

With CI (best from options) • Self-reported

93.0%

AA-LCR

Self-reported by the model provider • Self-reported

66.9%

AndroidWorld_SR

Self-reported by the model provider • Self-reported

66.4%

BabyVision

With CI (highest reported variant) • Self-reported

40.2%

BFCL-V4

Self-reported by the model provider • Self-reported

72.2%

BrowseComp

Self-reported by the model provider • Self-reported

63.8%

BrowseComp-zh

Self-reported by the model provider • Self-reported

69.9%

CC-OCR

Self-reported by the model provider • Self-reported

81.8%

C-Eval

Self-reported by the model provider • Self-reported

91.9%

CharXiv-R

RQ • Self-reported

77.2%

CodeForces

Elo Rating • Self-reported

85.1%

DeepPlanning

Self-reported by the model provider • Self-reported

24.1%

DynaMath

Self-reported by the model provider • Self-reported

85.9%

EmbSpatialBench

Self-reported by the model provider • Self-reported

83.9%

ERQA

Self-reported by the model provider • Self-reported

62.0%

FullStackBench en

Self-reported by the model provider • Self-reported

62.6%

FullStackBench zh

Self-reported by the model provider • Self-reported

58.7%

Global PIQA

Self-reported by the model provider • Self-reported

88.4%

Hallusion Bench

Self-reported by the model provider • Self-reported

67.6%

HMMT 2025

February 2025 • Self-reported

91.4%

HMMT25

November 2025 • Self-reported

90.3%

Humanity's Last Exam

With tools • Self-reported

47.5%

Hypersim

Self-reported by the model provider • Self-reported

12.7%

IFBench

Self-reported by the model provider • Self-reported

76.1%

Include

Self-reported by the model provider • Self-reported

82.8%

LingoQA

Self-reported by the model provider • Self-reported

80.8%

LiveCodeBench v6

Self-reported by the model provider • Self-reported

78.9%

LongBench v2

Self-reported by the model provider • Self-reported

60.2%

LVBench

Self-reported by the model provider • Self-reported

74.4%

MathVision

Self-reported by the model provider • Self-reported

86.2%

MathVista-Mini

Self-reported by the model provider • Self-reported

87.4%

MAXIFE

Self-reported by the model provider • Self-reported

87.9%

MedXpertQA

MM • Self-reported

67.3%

MLVU

Self-reported by the model provider • Self-reported

87.3%

MMBench-V1.1

EN_V1.1_dev • Self-reported

92.8%

MMLongBench-Doc

Self-reported by the model provider • Self-reported

59.0%

MMLU-Pro

Self-reported by the model provider • Self-reported

86.7%

MMLU-ProX

Self-reported by the model provider • Self-reported

82.2%

MMMLU

Self-reported by the model provider • Self-reported

86.7%

MMMU-Pro

Self-reported by the model provider • Self-reported

76.9%

MMStar

Self-reported by the model provider • Self-reported

82.9%

MMVU

Self-reported by the model provider • Self-reported

74.7%

Multi-Challenge

Self-reported by the model provider • Self-reported

61.5%

MVBench

Self-reported by the model provider • Self-reported

76.6%

NOVA-63

Self-reported by the model provider • Self-reported

58.6%

Nuscene

Self-reported by the model provider • Self-reported

15.4%

OCRBench

Self-reported by the model provider • Self-reported

92.1%

ODinW

13 • Self-reported

44.5%

OJBench

Self-reported by the model provider • Self-reported

39.5%

OmniDocBench 1.5

Self-reported by the model provider • Self-reported

89.8%

OSWorld-Verified

Self-reported by the model provider • Self-reported

58.0%

PMC-VQA

Self-reported by the model provider • Self-reported

63.3%

PolyMATH

Self-reported by the model provider • Self-reported

68.9%

RealWorldQA

Self-reported by the model provider • Self-reported

85.1%

RefCOCO-avg

Self-reported by the model provider • Self-reported

91.3%

RefSpatialBench

Self-reported by the model provider • Self-reported

69.3%

ScreenSpot Pro

Self-reported by the model provider • Self-reported

70.4%

Seal-0

Self-reported by the model provider • Self-reported

44.1%

SimpleVQA

Self-reported by the model provider • Self-reported

61.7%

SlakeVQA

Self-reported by the model provider • Self-reported

81.6%

SUNRGBD

Self-reported by the model provider • Self-reported

36.2%

SuperGPQA

Self-reported by the model provider • Self-reported

67.1%

t2-bench

Self-reported by the model provider • Self-reported

79.5%

Terminal-Bench 2.0

Self-reported by the model provider • Self-reported

49.4%

TIR-Bench

With CI (highest reported variant) • Self-reported

53.2%

VideoMME w/o sub.

Without subtitles • Self-reported

83.9%

VideoMME w sub.

With subtitles • Self-reported

87.3%

VideoMMMU

Self-reported by the model provider • Self-reported

82.0%

VITA-Bench

Self-reported by the model provider • Self-reported

33.6%

WideSearch

Self-reported by the model provider • Self-reported

60.5%

WMT24++

Self-reported by the model provider • Self-reported

78.3%

ZEROBench

Self-reported by the model provider • Self-reported

9.0%

ZEROBench-Sub

Self-reported by the model provider • Self-reported

36.2%

License & Metadata

License

apache-2.0

Announcement Date

March 1, 2026

Last Updated

March 21, 2026

Articles about Qwen3.5 122B A10B

The Two Loops: How China's Open-Source AI Strategy Is Outpacing America

A new USCC report warns that China's open AI models now dominate global downloads. 80% of US startups use Chinese models. Washington is scrambling.

March 25, 2026

9 min

Alibaba Doubles Down on Open Source at ModelScope DevCon

Alibaba confirmed its commitment to open-sourcing Qwen, Wan, and new model families at ModelScope DevCon in Nanjing. Here's what was announced.

March 24, 2026

2 min

Qwen 3.5 Is Alibaba's Bid to Win the Agentic AI Era

Alibaba's Qwen 3.5 family uses extreme MoE efficiency to beat models 7x its size. The flagship is now live on Arena — and claims to outperform GPT-5.2.

March 23, 2026

7 min

ik_llama.cpp Delivers 26x Faster Prompt Processing for Qwen 3.5

A new optimized C++ inference engine achieves 26x speedup on Qwen 3.5 prompt processing, a major win for local AI deployment.

March 22, 2026

2 min

Similar Models

All Models

Qwen3 Max Thinking

Alibaba

1.0T

Best score:0.9 (GPQA)

Released:Feb 2026

Price:$1.20/1M tokens

Qwen3-235B-A22B-Instruct-2507

Alibaba

235.0B

Best score:0.8 (GPQA)

Released:Jul 2025

Price:$0.15/1M tokens

Qwen3 235B A22B

Alibaba

235.0B

Best score:0.9 (MMLU)

Released:Apr 2025

Price:$0.20/1M tokens

Qwen3-235B-A22B-Thinking-2507

Alibaba

235.0B

Best score:0.8 (GPQA)

Released:Jul 2025

Price:$0.30/1M tokens

Qwen3-Coder 480B A35B Instruct

Alibaba

480.0B

Best score:0.8 (TAU)

Released:Jan 2025

Qwen3.5 9B

Alibaba

9.0B

Best score:0.8 (GPQA)

Released:Mar 2026

Qwen3.7 Max

Alibaba

Best score:0.9 (GPQA)

Released:May 2026

Price:$2.50/1M tokens

Qwen3.5 27B

Alibaba

27.0B

Best score:0.9 (GPQA)

Released:Mar 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.