Qwen3.5 27B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

Description

Qwen3.5-27B is a multimodal dense foundation model with 27 billion parameters. It combines strong reasoning, coding, multilingual, long-context, and visual understanding performance in a production-friendly open-weight package with a native 262K context window.

Release Date

2026-02-24

Parameters

27.0B

Context Length

262K

Modalities

image, text, video

Capability Radar

general

coding

reasoning

scienceest.

agents

multimodal

Science uses a reasoning proxy when dedicated science benchmarks are unavailable.

Rankings

Domain	#Rank	Score	Source
Agents & Tools	51	57.0	LS
Code Ranking	84	65.0	AA
General Ranking	43	80.0	AA
Multimodal Ranking	60	70.0	LS
Reasoning	54	67.0	LS
Science	63	68.0	AA

Benchmark Scores (LLM Stats)

3d

SUNRGBD

0.35 / 100SR

Hypersim

0.13 / 100SR

Agents

t2-bench

79.0%SR

BFCL-V4

68.5%SR

AndroidWorld_SR

64.2%SR

WideSearch

61.1%SR

BrowseComp

61.0%SR

FullStackBench en

60.1%SR

TIR-Bench

59.8%SR

FullStackBench zh

57.4%SR

OSWorld-Verified

56.2%SR

VITA-Bench

41.9%SR

Terminal-Bench 2.0

41.6%SR

DeepPlanning

22.6%SR

Biology

GPQA

85.5%SR

Chemistry

SuperGPQA

65.6%SR

Code

SWE-Bench Verified

72.4%SR

Communication

Multi-Challenge

60.8%SR

Embodied

EmbSpatialBench

0.84 / 100SR

Finance

MMLU-Pro

86.1%SR

MMLU-ProX

82.2%SR

General

IFEval

95.0%SR

MMLU-Redux

93.2%SR

C-Eval

90.5%SR

MAXIFE

88.0%SR

Global PIQA

87.5%SR

MMMLU

85.9%SR

MMMU

82.3%SR

Include

81.6%SR

MMStar

81.0%SR

LiveCodeBench v6

80.7%SR

IFBench

76.5%SR

MMMU-Pro

75.0%SR

LongBench v2

60.6%SR

NOVA-63

58.1%SR

SimpleVQA

0.56 / 100SR

Grounding

RefCOCO-avg

0.91 / 100SR

ScreenSpot Pro

70.3%SR

RefSpatialBench

0.68 / 100SR

Healthcare

VideoMMMU

82.3%SR

SlakeVQA

80.0%SR

MedXpertQA

62.4%SR

PMC-VQA

62.4%SR

Image To Text

OCRBench

89.4%SR

Language

LingoQA

82.0%SR

WMT24++

77.6%SR

Long Context

MLVU

85.9%SR

LVBench

73.6%SR

AA-LCR

66.1%SR

MMLongBench-Doc

0.60 / 100SR

Math

HMMT 2025

92.0%SR

HMMT25

89.8%SR

MathVista-Mini

87.8%SR

DynaMath

87.7%SR

MathVision

86.0%SR

CodeForces

0.81 / 3000SR

PolyMATH

71.2%SR

Humanity's Last Exam

48.5%SR

Multimodal

VLMsAreBlind

96.9%SR

93.7%SR

AI2D

92.9%SR

MMBench-V1.1

92.6%SR

OmniDocBench 1.5

88.9%SR

VideoMME w sub.

87.0%SR

VideoMME w/o sub.

82.8%SR

CC-OCR

81.0%SR

CharXiv-R

79.5%SR

MVBench

74.6%SR

MMVU

73.3%SR

BabyVision

44.6%SR

ZEROBench-Sub

0.36 / 100SR

Nuscene

15.2%SR

ZEROBench

0.10 / 100SR

Reasoning

CountBench

0.98 / 100SR

Hallusion Bench

70.0%SR

BrowseComp-zh

62.1%SR

ERQA

60.5%SR

Seal-0

47.2%SR

OJBench

40.1%SR

Spatial Reasoning

RealWorldQA

83.7%SR

Vision

ODinW

41.1%SR

AA Evaluation Indices

Intelligence Index

42.1

Coding Index

34.9

Tau2

0.9

Gpqa

0.9

Ifbench

0.8

Lcr

0.7

Scicode

0.4

Terminalbench Hard

0.3

Hle

0.2

LLM Stats Category Scores

Biology

Instruction Following

Structured Output

Text-to-image

Video

Chemistry

Embodied

Finance

General

Grounding

Image To Text

Language

Legal

Math

Physics

Spatial Reasoning

Vision

Economics

Frontend Development

Healthcare

Long Context

Multimodal

Reasoning

Tool Calling

Agents

Code

Communication

Spatial

Pricing

Input Price$0.3 / 1M tokens

Output Price$2.4 / 1M tokens

Blended Price (3:1)$0.825 / 1M tokens

Speed

Tokens/sec87.6 tokens/s

Time to First Token1.40s

Time to Answer24.23s

Available Providers

(LS internal units)

Provider	Input Price	Output Price
Novita	300K	2.4M

External Sources

LLM Stats