Qwen3.5-35B-A3B

Qwen3.5-35B-A3B places in the top three on 5 benchmarks.

Qwen Released February 24th, 2026 Compare with others

118

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Qwen Feb 2026

Qwen3.5-35B-A3B

Open weights

144.0 #102 of 256

Epoch Capabilities Index

Self-Reported Index 144.6 #136 of 404

MMMU-Pro 75.1% #40 /90
IFBench 70.2% #44 /80
AA-LCR (Artificial Analysis Long Context Reasoning) 58.5% #24 /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
BOUQuET	31.4	#2 of 18	Official leaderboard
MMLU-Pro	85.3%	#23 of 167	Lab model card
IFEval	91.9%	#21 of 135	Lab model card
QIMMA	61.2%	#6 of 25	Official leaderboard
HallusionBench	67.9%	#8 of 31	Lab model card
Judgemark v2.1	81.4	#12 of 42	Official leaderboard
MMLU-Redux	93.3%	#13 of 45	Lab model card
SEA-HELM	73.5	#22 of 71	Official leaderboard
Humanity's Last Exam	47.4%	#52 of 154	Lab model card
MMLU-ProX	81%	#9 of 23	Lab model card
KINA	35.4%	#15 of 37	Official leaderboard
LMArena (Elo)	1,394	#138 of 335	Official leaderboard
C-Eval	90.2%	#19 of 46	Lab model card
INCLUDE	79.7%	#13 of 31	Lab model card
MMMLU	85.2%	#34 of 71	Lab model card
LongBench v2	59%	#17 of 35	Lab model card
SuperGPQA	63.4%	#17 of 35	Lab model card
MultiChallenge	60%	#21 of 42	Lab model card
IFBench	70.2%	#44 of 80	Lab model card
Vectara Hallucination Leaderboard (2025 dataset)	10.5%	#64 of 104	Official leaderboard
AA-LCR (Artificial Analysis Long Context Reasoning)	58.5%	#24 of 38	Lab model card
MAXIFE	86.6%	#7 of 11	Lab model card
NOVA-63	57.1%	#7 of 11	Lab model card
SimpleVQA	58.3%	#21 of 30	Lab model card
Global PIQA	86.6%	#10 of 14	Lab model card
Judgemark v4	40.5	#30 of 40	Official leaderboard
Longform Creative Writing (EQ-Bench)	44.5	#93 of 119	Official leaderboard
WideSearch	57.1%	#19 of 19	Lab model card
Seal-0	41.4%	#7 of 7	Lab model card

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	83.5%	#85 of 229	Epoch AI run
PolyMath	64.4%	#9 of 24	Lab model card
HMMT 2025	89%	#26 of 54	Lab model card
DTBench	80%	#93 of 180	Official leaderboard
OTIS Mock AIME 2024-2025	70%	#111 of 212	Epoch AI run
MathArena (Apex, IMO, USAMO)	4.17%	#28 of 47	Official leaderboard
HMMT Nov 2025	89.2%	#17 of 28	Lab model card
AIME 2026	93.3%	#23 of 31	Official leaderboard
HMMT 2026	81.8%	#26 of 31	Official leaderboard
MathArena Apex Shortlist	44.7%	#33 of 38	Official leaderboard

Coding

Benchmark	Score	Rank	Source
LiveCodeBench v6	74.6%	#45 of 90	Lab model card
SWE-bench Verified	69.2%	#85 of 155	Lab model card
Codeforces rating	2,028	#26 of 47	Lab model card
OJBench	36%	#9 of 16	Lab model card
FullStackBench	56.5%	#7 of 8	Lab model card
Code Arena	1,253	#115 of 128	Official leaderboard
ALE-Bench	648	#124 of 126	Official leaderboard
SWE-rebench	17.1%	#13 of 13	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
GameWorld	43.8%	#6 of 22	Official leaderboard
AndroidWorld	71.1%	#8 of 23	Lab model card
tau2-bench	81.2%	#30 of 79	Lab model card
BFCL V4	67.3%	#12 of 30	Lab model card
SnakeBench	25	#78 of 188	Official leaderboard
BrowseComp-ZH	69.5%	#8 of 19	Lab model card
ScreenSpot-Pro	68.6%	#16 of 32	Lab model card
LisanBench	447.3	#70 of 133	Official leaderboard
LLM Chess	-26	#66 of 123	Official leaderboard
Toolathlon	28.7%	#32 of 50	Official leaderboard
VITA-Bench	31.9%	#11 of 16	Lab model card
DeepPlanning	22.8%	#7 of 10	Lab model card
BrowseComp	61%	#57 of 81	Lab model card
Terminal-Bench 2.0	40.5%	#51 of 63	Lab model card
OSWorld-Verified	54.5%	#34 of 41	Lab model card
MyPCBench	7.60%	#8 of 9	Official leaderboard
RuneBench	0.75	#76 of 76	Official leaderboard
AutoResearchBench	1.94%	#13 of 13	Official leaderboard
iOSWorld	12.8%	#6 of 6	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
MedXpertQA-MM	61.4%	#8 of 21	Lab model card
LMBuild	47%	#9 of 18	Official leaderboard
PMC-VQA	62%	#6 of 10	Lab model card
SLAKE	78.7%	#7 of 10	Lab model card
Vending-Bench 2	$463	#57 of 67	Official leaderboard
LEXam-hard	23.9%	#13 of 15	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
CountBench	97.8%	#2 of 28	Lab model card
VLMs Are Blind	97%	#2 of 14	Lab model card
MathVista	86.2%	#11 of 70	Lab model card
MMMU	81.4%	#17 of 103	Lab model card
MMStar	81.9%	#6 of 36	Lab model card
OCRBench	91%	#6 of 34	Lab model card
RefAdv-S	66.7%	#4 of 19	Official leaderboard
Hypersim	13.1	#2 of 9	Lab model card
DynaMath	85%	#6 of 23	Lab model card
RealWorldQA	84.1%	#11 of 38	Lab model card
AI2D	92.6%	#16 of 52	Lab model card
Video-MME	86.6%	#16 of 51	Lab model card
MMLongBench-Doc	59.5%	#6 of 19	Lab model card
MMBench-EN v1.1	91.5%	#8 of 23	Lab model card
ZeroBench	34.1%	#11 of 31	Lab model card
OCRBench v2	65.3%	#5 of 14	Official leaderboard
MATH-Vision (MathVision)	83.9%	#19 of 51	Lab model card
ERQA	64.8%	#14 of 36	Lab model card
MMVU	72.3%	#11 of 26	Lab model card
MVBench	74.8%	#10 of 23	Lab model card
MMMU-Pro	75.1%	#40 of 90	Lab model card
VideoMME v2 (with subtitles)	25.8%	#14 of 31	Official leaderboard
MLVU	85.6%	#10 of 22	Lab model card
V*	92.7%	#8 of 17	Lab model card
RefSpatial-Bench	63.5%	#8 of 16	Lab model card
GeoGuesser environment (FineEnvs)	0.33	#7 of 14	Official leaderboard
Nuscene	14.6	#4 of 8	Lab model card
TIR-Bench	55.5%	#3 of 6	Lab model card
LVBench	71.4%	#17 of 31	Lab model card
RefCOCO / RefCOCO+ / RefCOCOg	89.2%	#16 of 29	Lab model card
CharXiv	77.5%	#44 of 75	Lab model card
Video-MMMU	80.4%	#22 of 36	Lab model card
EmbSpatialBench	83.1%	#10 of 16	Lab model card
CC-OCR	80.7%	#9 of 14	Lab model card
SUNRGBD	33.4	#6 of 9	Lab model card
ODinW-13	42.6%	#9 of 13	Lab model card
LingoQA	79.2%	#7 of 10	Lab model card
BabyVision	38.4%	#21 of 26	Lab model card
AgentWorldBench	47.7	#12 of 13	Official leaderboard
CiteVQA	10.7%	#15 of 16	Official leaderboard

Safety and security

Benchmark	Score		Rank	Source
SusVibes	18.3%		#4 of 10	Official leaderboard
nocot-bench (NCRI)	60.1		#137 of 185	Official leaderboard

Other

Benchmark	Score	Rank	Source
Conceptual Reasoning Index	42.1	#94 of 151	Official leaderboard
LMCA	29.5%	#92 of 147	Official leaderboard
ACCoRD	55.3	#106 of 152	Official leaderboard
Political Science LLM Benchmark	65%	#30 of 42	Official leaderboard

Its lab's numbers put it at 144.6, in line with its ECI

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 143.6 to 145.5. How we compute it.

CountBench 97.8
VLMs Are Blind 97
MMLU-Redux 93.3
V* 92.7
AI2D 92.6
IFEval 91.9
MMBench-EN v1.1 91.5
OCRBench 91
C-Eval 90.2
HMMT Nov 2025 89.2
RefCOCO / RefCOCO+ / RefCOCOg 89.2
HMMT 2025 89
Global PIQA 86.6
MAXIFE 86.6
Video-MME 86.6
MathVista 86.2
MLVU 85.6
MMLU-Pro 85.3
MMMLU 85.2
DynaMath 85
GPQA Diamond 84.2
RealWorldQA 84.1
MATH-Vision (MathVision) 83.9
EmbSpatialBench 83.1
MMStar 81.9
MMMU 81.4
tau2-bench 81.2
MMLU-ProX 81
CC-OCR 80.7
Video-MMMU 80.4
INCLUDE 79.7
LingoQA 79.2
SLAKE 78.7
CharXiv 77.5
MMMU-Pro 75.1
MVBench 74.8
LiveCodeBench v6 74.6
MMVU 72.3
LVBench 71.4
AndroidWorld 71.1
IFBench 70.2
BrowseComp-ZH 69.5
SWE-bench Verified 69.2
ScreenSpot-Pro 68.6
HallusionBench 67.9
BFCL V4 67.3
PolyMath 64.4
SuperGPQA 63.4
PMC-VQA 62
MedXpertQA-MM 61.4
BrowseComp 61
MultiChallenge 60
MMLongBench-Doc 59.5
LongBench v2 59
AA-LCR (Artificial Analysis Long Context Reasoning) 58.5
SimpleVQA 58.3
WideSearch 57.1
NOVA-63 57.1
FullStackBench 56.5
OSWorld-Verified 54.5
Humanity's Last Exam 47.4
ODinW-13 42.6
Terminal-Bench 2.0 40.5
BabyVision 38.4
OJBench 36
ZeroBench 34.1
VITA-Bench 31.9
DeepPlanning 22.8

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
GPQA Diamond	84.2% GPQA Diamond	83.5% Epoch AI run	+0.7 pts
MATH-Vision (MathVision)	83.9% MathVision	83.9% Leaderboard	Same
MMLU-Pro	85.3% MMLU-Pro	85.3% Leaderboard	Same

Qwen's models over time

Qwen's models in color, against every other model on the Epoch Capabilities Index, by release date.

Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)

Its launch post

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

Qwen3.5-35B-A3B model card (Hugging Face), Qwen3.5 medium series Feb 24, 2026

MMLU-Pro 85.3
MMLU-Redux 93.3
C-Eval 90.2
SuperGPQA 63.4
IFEval 91.9
IFBench 70.2
MultiChallenge 60
AA-LCR 58.5
LongBench v2 59
GPQA Diamond 84.2
HMMT Feb 25 89
HMMT Nov 25 89.2
SWE-bench Verified 69.2
Terminal Bench 2 40.5
LiveCodeBench v6 74.6
CodeForces 2028
OJBench 36
Mean of FullStackBench en 58.1 / FullStackBench zh 55.0 56.5
BFCL-V4 67.3
TAU2-Bench 81.2
VITA-Bench 31.9
DeepPlanning 22.8
Browsecomp 61
Browsecomp-zh 69.5
WideSearch 57.1
Seal-0 41.4
MMMLU 85.2
MMLU-ProX 81
NOVA-63 57.1
INCLUDE 79.7
Global PIQA 86.6
PolyMATH 64.4
WMT24++ 76.3
MAXIFE 86.6
MMMU 81.4
MMMU-Pro 75.1
MathVision 83.9
Mathvista(mini) 86.2
DynaMath 85
Best of ZEROBench / ZEROBench_sub 34.1 34.1
VlmsAreBlind 97
BabyVision 38.4 / 29.6 (with CI / without CI) 38.4
RealWorldQA 84.1
MMStar 81.9
MMBench EN-DEV-v1.1 91.5
SimpleVQA 58.3
HallusionBench 67.9
OmniDocBench1.5 89.3
CharXiv(RQ) 77.5
MMLongBench-Doc 59.5
CC-OCR 80.7
AI2D_TEST 92.6
OCRBench 91
ERQA 64.8
CountBench 97.8
RefCOCO(avg) 89.2
ODInW13 42.6
EmbSpatialBench 83.1
RefSpatialBench 63.5
LingoQA 79.2
Hypersim 13.1
SUNRGBD 33.4
Nuscene 14.6
Best of VideoMME(w sub.) / VideoMME(w/o sub.) 86.6 86.6
VideoMMMU 80.4
MLVU 85.6
MVBench 74.8
LVBench 71.4
MMVU 72.3
ScreenSpot Pro 68.6
OSWorld-Verified 54.5
AndroidWorld 71.1
TIR-Bench 55.5 / 38.0 (with CI / without CI) 55.5
V* 92.7 / 89.5 (with CI / without CI) 92.7
SLAKE 78.7
PMC-VQA 62
MedXpertQA-MM 61.4
HLE w/ CoT 22.4; HLE w/ tool 47.4 47.4