Qwen3.5-27B

Qwen3.5-27B ranks first on 3 of the 120 benchmarks with five or more models scored.

Qwen Released February 24th, 2026 Compare with others

120

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Qwen Feb 2026

Qwen3.5-27B

Open weights

146.0 #122 of 404

Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.

MMMU-Pro 75% #41 /90
IFBench 76.5% #19 /80
AA-LCR (Artificial Analysis Long Context Reasoning) 66.1% #16 /38

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
IFEval	95%	#7 of 135	Lab model card
MMLU-Pro	86.1%	#19 of 167	Lab model card
IFMTBench	81.1%	#1 of 8	Official leaderboard
HallusionBench	70%	#4 of 31	Lab model card
QIMMA	62%	#5 of 25	Official leaderboard
STALE	31.3%	#2 of 10	Official leaderboard
Judgemark v2.1	84.3	#9 of 42	Official leaderboard
BOUQuET	31.2	#4 of 18	Official leaderboard
Global PIQA	78.5%	#10 of 44	Official leaderboard
IFBench	76.5%	#19 of 80	Lab model card
SEA-HELM	74.2	#17 of 71	Official leaderboard
KINA	39.4%	#10 of 37	Official leaderboard
Humanity's Last Exam	48.5%	#45 of 154	Lab model card
MMLU-ProX	82.2%	#7 of 23	Lab model card
INCLUDE	81.6%	#10 of 31	Lab model card
MMLU-Redux	93.2%	#15 of 45	Lab model card
SuperGPQA	65.6%	#12 of 35	Lab model card
LMArena (Elo)	1,409	#123 of 335	Official leaderboard
C-Eval	90.5%	#17 of 46	Lab model card
MMMLU	85.9%	#27 of 71	Lab model card
LongBench v2	60.6%	#14 of 35	Lab model card
CulturalRiddles	28%	#8 of 20	Official leaderboard
AA-LCR (Artificial Analysis Long Context Reasoning)	66.1%	#16 of 38	Lab model card
MultiChallenge	60.8%	#19 of 42	Lab model card
MAXIFE	88%	#5 of 11	Lab model card
NOVA-63	58.1%	#5 of 11	Lab model card
Judgemark v4	60.5	#19 of 40	Official leaderboard
Longform Creative Writing (EQ-Bench)	59.0	#65 of 119	Official leaderboard
Seal-0	47.2%	#4 of 7	Lab model card
Extended NYT Connections	47.9%	#72 of 103	Official leaderboard
SimpleVQA	56%	#23 of 30	Lab model card
Vectara Hallucination Leaderboard (2025 dataset)	12.1%	#81 of 104	Official leaderboard
WideSearch	61.1%	#16 of 19	Lab model card

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	85.5%	#75 of 302	Lab model card
PolyMath	71.2%	#6 of 24	Lab model card
HMMT 2025	92%	#20 of 54	Lab model card
DTBench	82.4%	#82 of 180	Official leaderboard
HMMT Nov 2025	89.8%	#16 of 28	Lab model card
MathArena (Apex, IMO, USAMO)	2.08%	#33 of 47	Official leaderboard
MathArena Apex Shortlist	52.1%	#29 of 38	Official leaderboard
AIME 2026	91.7%	#26 of 31	Official leaderboard
HMMT 2026	81.1%	#27 of 31	Official leaderboard

Coding

Benchmark	Score	Rank	Source
LiveCodeBench v6	80.7%	#30 of 90	Lab model card
SWE-bench Verified	72.4%	#62 of 155	Lab model card
OJBench	40.1%	#7 of 16	Lab model card
Codeforces rating	1,899	#31 of 47	Lab model card
FullStackBench	58.8%	#6 of 8	Lab model card
WebDev Arena	1,357	#80 of 105	arena.ai, via Epoch AI
Code Arena	1,357	#101 of 128	Official leaderboard
ALE-Bench	882	#116 of 126	Official leaderboard
FeatureBench	34.8%	#15 of 16	Official leaderboard
FeatureBench Lite	34.8%	#15 of 16	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
SnakeBench	31	#31 of 188	Official leaderboard
BFCL V4	68.5%	#10 of 30	Lab model card
VITA-Bench	41.9%	#6 of 16	Lab model card
GameWorld	41.8%	#9 of 22	Official leaderboard
ScreenSpot-Pro	70.3%	#14 of 32	Lab model card
tau2-bench	79%	#42 of 79	Lab model card
OpenComputer	32.3%	#4 of 7	Official leaderboard
Toolathlon	31.5%	#30 of 50	Official leaderboard
AndroidWorld	64.2%	#15 of 23	Lab model card
BrowseComp-ZH	62.1%	#13 of 19	Lab model card
BrowseComp	61%	#56 of 81	Lab model card
Terminal-Bench 2.0	41.6%	#49 of 63	Lab model card
ScienceBoard	5.92%	#12 of 15	Official leaderboard
DeepPlanning	22.6%	#8 of 10	Lab model card
OSWorld-Verified	56.2%	#33 of 41	Lab model card
LiveClawBench	31.6%	#17 of 17	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
MedXpertQA-MM	62.4%	#7 of 21	Lab model card
LMBuild	55.3%	#6 of 18	Official leaderboard
SLAKE	80%	#4 of 10	Lab model card
PMC-VQA	62.4%	#5 of 10	Lab model card
MMEnterprise-Bench	50.3%	#5 of 9	Official leaderboard
Vending-Bench 2	$202	#59 of 67	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
CountBench	97.8%	#1 of 28	Lab model card
RefAdv-S	67.3%	#1 of 19	Official leaderboard
MathVista	87.8%	#5 of 70	Lab model card
DynaMath	87.7%	#2 of 23	Lab model card
MMMU	82.3%	#11 of 103	Lab model card
GSM8K-V	39.9%	#4 of 24	Official leaderboard
MMBench-EN v1.1	92.6%	#5 of 23	Lab model card
MMStar	81%	#8 of 36	Lab model card
ZeroBench	36.2%	#7 of 31	Lab model card
OCRBench	89.4%	#8 of 34	Lab model card
EmbSpatialBench	84.5%	#4 of 16	Lab model card
VideoMME v2 (with subtitles)	31.4%	#8 of 31	Official leaderboard
MMLongBench-Doc	60.2%	#5 of 19	Lab model card
OVO-S-Bench	51.7%	#4 of 15	Official leaderboard
AI2D	92.9%	#14 of 52	Lab model card
Video-MME	87%	#14 of 51	Lab model card
VLMs Are Blind	96.9%	#4 of 14	Lab model card
LingoQA	82%	#3 of 10	Lab model card
MATH-Vision (MathVision)	86%	#16 of 51	Lab model card
Hypersim	13.0	#3 of 9	Lab model card
TIR-Bench	59.8%	#2 of 6	Lab model card
MMVU	73.3%	#9 of 26	Lab model card
V*	93.7%	#6 of 17	Lab model card
RefSpatial-Bench	67.7%	#6 of 16	Lab model card
Nuscene	15.2	#3 of 8	Lab model card
RealWorldQA	83.7%	#15 of 38	Lab model card
MLVU	85.9%	#9 of 22	Lab model card
RefCOCO / RefCOCO+ / RefCOCOg	90.9%	#12 of 29	Lab model card
GeoGuesser environment (FineEnvs)	0.33	#6 of 14	Official leaderboard
SUNRGBD	35.4	#4 of 9	Lab model card
MMMU-Pro	75%	#41 of 90	Lab model card
MVBench	74.6%	#11 of 23	Lab model card
LVBench	73.6%	#15 of 31	Lab model card
Vision Arena	1,216	#67 of 136	Official leaderboard
Video-MMMU	82.3%	#18 of 36	Lab model card
CharXiv	79.5%	#38 of 75	Lab model card
ERQA	60.5%	#19 of 36	Lab model card
CC-OCR	81%	#8 of 14	Lab model card
CiteVQA	17.3%	#11 of 16	Official leaderboard
BabyVision	44.6%	#18 of 26	Lab model card
ODinW-13	41.1%	#11 of 13	Lab model card
BVB	50.7	#45 of 45	Official leaderboard

Safety and security

Benchmark	Score		Rank	Source
nocot-bench (NCRI)	66.6		#114 of 185	Official leaderboard

Other

Benchmark	Score	Rank	Source
Conceptual Reasoning Index	46.6	#80 of 151	Official leaderboard
LMCA	34%	#82 of 147	Official leaderboard
ACCoRD	60.5	#90 of 152	Official leaderboard

Its lab's numbers put it at 146.0, number 122 of 404

The Self-Reported Index fits one score to the 68 capability results Qwen reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 145.0 to 147.1. How we compute it.

CountBench 97.8
VLMs Are Blind 96.9
IFEval 95
V* 93.7
MMLU-Redux 93.2
AI2D 92.9
MMBench-EN v1.1 92.6
HMMT 2025 92
RefCOCO / RefCOCO+ / RefCOCOg 90.9
C-Eval 90.5
HMMT Nov 2025 89.8
OCRBench 89.4
MAXIFE 88
MathVista 87.8
DynaMath 87.7
Global PIQA 87.5
Video-MME 87
MMLU-Pro 86.1
MATH-Vision (MathVision) 86
MMMLU 85.9
MLVU 85.9
GPQA Diamond 85.5
EmbSpatialBench 84.5
RealWorldQA 83.7
MMMU 82.3
Video-MMMU 82.3
MMLU-ProX 82.2
LingoQA 82
INCLUDE 81.6
MMStar 81
CC-OCR 81
LiveCodeBench v6 80.7
SLAKE 80
CharXiv 79.5
tau2-bench 79
IFBench 76.5
MMMU-Pro 75
MVBench 74.6
LVBench 73.6
MMVU 73.3
SWE-bench Verified 72.4
PolyMath 71.2
ScreenSpot-Pro 70.3
HallusionBench 70
BFCL V4 68.5
AA-LCR (Artificial Analysis Long Context Reasoning) 66.1
SuperGPQA 65.6
AndroidWorld 64.2
PMC-VQA 62.4
MedXpertQA-MM 62.4
BrowseComp-ZH 62.1
WideSearch 61.1
BrowseComp 61
MultiChallenge 60.8
LongBench v2 60.6
MMLongBench-Doc 60.2
FullStackBench 58.8
NOVA-63 58.1
OSWorld-Verified 56.2
SimpleVQA 56
Humanity's Last Exam 48.5
BabyVision 44.6
VITA-Bench 41.9
Terminal-Bench 2.0 41.6
ODinW-13 41.1
OJBench 40.1
ZeroBench 36.2
DeepPlanning 22.6

Its lab's numbers next to 3 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap
Global PIQA	87.5% Global PIQA	78.5% Leaderboard, exact_match,strict_match; instruct	+9.0 pts
MATH-Vision (MathVision)	86% MathVision	86% Leaderboard	Same
MMLU-Pro	86.1% MMLU-Pro	86.1% Leaderboard	Same

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Qwen3.5-27B model card (Hugging Face), Qwen3.5 medium series Feb 24, 2026

MMLU-Pro 86.1
MMLU-Redux 93.2
C-Eval 90.5
SuperGPQA 65.6
IFEval 95
IFBench 76.5
MultiChallenge 60.8
AA-LCR 66.1
LongBench v2 60.6
GPQA Diamond 85.5
HMMT Feb 25 92
HMMT Nov 25 89.8
SWE-bench Verified 72.4
Terminal Bench 2 41.6
LiveCodeBench v6 80.7
CodeForces 1899
OJBench 40.1
Mean of FullStackBench en 60.1 / FullStackBench zh 57.4 58.8
BFCL-V4 68.5
TAU2-Bench 79
VITA-Bench 41.9
DeepPlanning 22.6
Browsecomp 61
Browsecomp-zh 62.1
WideSearch 61.1
Seal-0 47.2
MMMLU 85.9
MMLU-ProX 82.2
NOVA-63 58.1
INCLUDE 81.6
Global PIQA 87.5
PolyMATH 71.2
WMT24++ 77.6
MAXIFE 88
MMMU 82.3
MMMU-Pro 75
MathVision 86
Mathvista(mini) 87.8
DynaMath 87.7
Best of ZEROBench / ZEROBench_sub 36.2 36.2
VlmsAreBlind 96.9
BabyVision 44.6 / 34.8 (with CI / without CI) 44.6
RealWorldQA 83.7
MMStar 81
MMBench EN-DEV-v1.1 92.6
SimpleVQA 56
HallusionBench 70
OmniDocBench1.5 88.9
CharXiv(RQ) 79.5
MMLongBench-Doc 60.2
CC-OCR 81
AI2D_TEST 92.9
OCRBench 89.4
ERQA 60.5
CountBench 97.8
RefCOCO(avg) 90.9
ODInW13 41.1
EmbSpatialBench 84.5
RefSpatialBench 67.7
LingoQA 82
Hypersim 13
SUNRGBD 35.4
Nuscene 15.2
Best of VideoMME(w sub.) / VideoMME(w/o sub.) 87.0 87
VideoMMMU 82.3
MLVU 85.9
MVBench 74.6
LVBench 73.6
MMVU 73.3
ScreenSpot Pro 70.3
OSWorld-Verified 56.2
AndroidWorld 64.2
TIR-Bench 59.8 / 42.3 (with CI / without CI) 59.8
V* 93.7 / 89.0 (with CI / without CI) 93.7
SLAKE 80
PMC-VQA 62.4
MedXpertQA-MM 62.4
HLE w/ CoT 24.3; HLE w/ tool 48.5 48.5