Qwen3.5-9B

Qwen3.5-9B ranks first on Hypersim.

Qwen Released February 24th, 2026 Compare with others

114

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Qwen Feb 2026

Qwen3.5-9B

Open weights

139.3 #127 of 256

Epoch Capabilities Index

Self-Reported Index 140.5 #174 of 404

MMMU-Pro 70.1% #50 /90
IFBench 64.5% #53 /80
Terminal-Bench 2.0 9.21% #42 /43

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
AgentIF (2026-10 setting)	65.3%	#4 of 24	Official leaderboard
IFEval	91.5%	#24 of 135	Lab model card
HallusionBench	69.3%	#6 of 31	Lab model card
QIMMA	59.1%	#8 of 25	Official leaderboard
MMLU-Pro	82.5%	#57 of 167	Lab model card
Global PIQA	73.5%	#18 of 44	Official leaderboard
SEA-HELM	69.1	#35 of 71	Official leaderboard
Judgemark v2.1	75.0	#21 of 42	Official leaderboard
BOUQuET	28.4	#9 of 18	Official leaderboard
IFMTBench	71.2%	#4 of 8	Official leaderboard
MMLU-Redux	91.1%	#23 of 45	Lab model card
INCLUDE	75.6%	#16 of 31	Lab model card
AA-LCR (Artificial Analysis Long Context Reasoning)	63%	#20 of 38	Lab model card
KINA	30.1%	#20 of 37	Official leaderboard
C-Eval	88.2%	#26 of 46	Lab model card
LongBench v2	55.2%	#20 of 35	Lab model card
MMMLU	81.2%	#42 of 71	Lab model card
SuperGPQA	58.2%	#21 of 35	Lab model card
STALE	14.6%	#6 of 10	Official leaderboard
MultiChallenge	54.5%	#27 of 42	Lab model card
MMLU-ProX	76.3%	#15 of 23	Lab model card
IFBench	64.5%	#53 of 80	Lab model card
MAXIFE	83.4%	#8 of 11	Lab model card
NOVA-63	55.9%	#8 of 11	Lab model card
HELM Arabic Enterprise	62.4	#17 of 23	Official leaderboard
HELM Arabic	73.1	#22 of 28	Official leaderboard
Judgemark v4	32.4	#33 of 40	Official leaderboard
Wer wird Millionär? LLM Benchmark	169.0	#64 of 77	Official leaderboard
SimpleVQA	51.2%	#26 of 30	Lab model card

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	79%	#100 of 229	Epoch AI run
PolyMath	57.3%	#12 of 24	Lab model card
PHYSICS-Verified	73.8%	#10 of 19	Official leaderboard
OTIS Mock AIME 2024-2025	61.7%	#124 of 212	Epoch AI run
DTBench	71.2%	#111 of 180	Official leaderboard
HMMT 2025	83.2%	#36 of 54	Lab model card
AIME 2026	92.5%	#24 of 31	Official leaderboard
HMMT Nov 2025	82.9%	#22 of 28	Lab model card
MathArena Apex Shortlist	29.8%	#35 of 38	Official leaderboard
HMMT 2026	71.2%	#29 of 31	Official leaderboard
MathArena (Apex, IMO, USAMO)	0.52%	#46 of 47	Official leaderboard

Coding

Benchmark	Score	Rank	Source
Aspire	25.6%	#1 of 2	Official leaderboard
LiveCodeBench v6	65.6%	#66 of 90	Lab model card
OJBench	29.2%	#12 of 16	Lab model card
SWE-bench Science	7.56%	#10 of 10	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
BFCL V4	66.1%	#14 of 30	Lab model card
tau2-bench	79.1%	#41 of 79	Lab model card
GameWorld	39.3%	#13 of 22	Official leaderboard
ScreenSpot-Pro	65.2%	#20 of 32	Lab model card
VITA-Bench	29.8%	#12 of 16	Lab model card
LLM Chess	-200	#94 of 123	Official leaderboard
AndroidWorld	57.8%	#19 of 23	Lab model card
AdaExploreBench	28.2%	#7 of 8	Official leaderboard
GAIA	13.9%	#57 of 64	Official leaderboard
DeepPlanning	18%	#9 of 10	Lab model card
OSWorld-Verified	41.8%	#38 of 41	Lab model card
PinchBench	34.8%	#46 of 48	Official leaderboard
Terminal-Bench 2.0	9.21%	#42 of 43	Official leaderboard
WitnessBench	0.11%	#12 of 12	Official leaderboard
MyPCBench	6.50%	#9 of 9	Official leaderboard
OpenComputer	7.80%	#7 of 7	Official leaderboard
Gym-Anything-RT	0%	#5 of 5	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
MedXpertQA-MM	49.9%	#12 of 21	Lab model card
SLAKE	79%	#6 of 10	Lab model card
PMC-VQA	57.9%	#7 of 10	Lab model card
MMEnterprise-Bench	43.1%	#7 of 9	Official leaderboard
CorpBench Work	62.4	#16 of 17	Official leaderboard
FoodTruck Bench	$-679	#59 of 59	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
Hypersim	13.5	#1 of 9	Lab model card
MathVista	85.7%	#14 of 70	Lab model card
CountBench	97.2%	#6 of 28	Lab model card
MMMU	78.4%	#25 of 103	Lab model card
MMStar	79.7%	#10 of 36	Lab model card
OCRBench	89.2%	#10 of 34	Lab model card
AI2D	90.2%	#18 of 52	Lab model card
DynaMath	83.6%	#8 of 23	Lab model card
VideoMME v2 (with subtitles)	26.2%	#13 of 31	Official leaderboard
ZeroBench	31.1%	#13 of 31	Lab model card
MMLongBench-Doc	57.7%	#8 of 19	Lab model card
VLMs Are Blind	93.7%	#6 of 14	Lab model card
Video-MME	84.5%	#22 of 51	Lab model card
MMBench-EN v1.1	90.1%	#10 of 23	Lab model card
RealWorldQA	80.3%	#17 of 38	Lab model card
MATH-Vision (MathVision)	78.9%	#23 of 51	Lab model card
GSM8K-V	27.8%	#11 of 24	Official leaderboard
We-Math	75.2%	#6 of 13	Lab model card
CC-OCR-Bench-V2	67.3%	#7 of 15	Official leaderboard
MLVU	84.4%	#11 of 22	Lab model card
RefCOCO / RefCOCO+ / RefCOCOg	89.7%	#15 of 29	Lab model card
ExtractBench	86.1%	#14 of 27	Official leaderboard
V*	90.1%	#9 of 17	Lab model card
MMMU-Pro	70.1%	#50 of 90	Lab model card
RefSpatial-Bench	58.5%	#9 of 16	Lab model card
GeoGuesser environment (FineEnvs)	0.33	#8 of 14	Official leaderboard
OVO-S-Bench	45.9%	#9 of 15	Official leaderboard
LingoQA	80.4%	#6 of 10	Lab model card
MVBench	74.4%	#14 of 23	Lab model card
ERQA	55.5%	#22 of 36	Lab model card
Nuscene	11.8	#5 of 8	Lab model card
LVBench	70%	#20 of 31	Lab model card
CharXiv	73%	#50 of 75	Lab model card
EmbSpatialBench	83%	#11 of 16	Lab model card
MMVU	67.8%	#18 of 26	Lab model card
Video-MMMU	78.9%	#26 of 36	Lab model card
CC-OCR	79.3%	#11 of 14	Lab model card
OCRBench v2	58.7%	#11 of 14	Official leaderboard
TIR-Bench	45.6%	#5 of 6	Lab model card
CiteVQA	11.1%	#14 of 16	Official leaderboard
ParseBench	31.9%	#40 of 44	Official leaderboard
MDPBench	65.7	#11 of 12	Official leaderboard
BabyVision	28.6%	#24 of 26	Lab model card

Safety and security

Benchmark	Score		Rank	Source
nocot-bench (NCRI)	52.4		#160 of 185	Official leaderboard

Other

Benchmark	Score	Rank	Source
LMCA	24.5%	#103 of 147	Official leaderboard
Conceptual Reasoning Index	35.1	#109 of 151	Official leaderboard
ACCoRD	50.2	#112 of 152	Official leaderboard

Its lab's numbers put it at 140.5, 1.1 points above its ECI

The Self-Reported Index fits one score to the 61 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 139.5 to 141.4. How we compute it.

CountBench 97.2
VLMs Are Blind 93.7
IFEval 91.5
MMLU-Redux 91.1
AI2D 90.2
MMBench-EN v1.1 90.1
V* 90.1
RefCOCO / RefCOCO+ / RefCOCOg 89.7
OCRBench 89.2
C-Eval 88.2
MathVista 85.7
Video-MME 84.5
MLVU 84.4
DynaMath 83.6
MAXIFE 83.4
HMMT 2025 83.2
Global PIQA 83.2
EmbSpatialBench 83
HMMT Nov 2025 82.9
MMLU-Pro 82.5
GPQA Diamond 81.7
MMMLU 81.2
LingoQA 80.4
RealWorldQA 80.3
MMStar 79.7
CC-OCR 79.3
tau2-bench 79.1
SLAKE 79
MATH-Vision (MathVision) 78.9
Video-MMMU 78.9
MMMU 78.4
MMLU-ProX 76.3
INCLUDE 75.6
We-Math 75.2
MVBench 74.4
CharXiv 73
MMMU-Pro 70.1
LVBench 70
HallusionBench 69.3
MMVU 67.8
BFCL V4 66.1
LiveCodeBench v6 65.6
ScreenSpot-Pro 65.2
IFBench 64.5
AA-LCR (Artificial Analysis Long Context Reasoning) 63
SuperGPQA 58.2
PMC-VQA 57.9
AndroidWorld 57.8
MMLongBench-Doc 57.7
PolyMath 57.3
NOVA-63 55.9
LongBench v2 55.2
MultiChallenge 54.5
SimpleVQA 51.2
MedXpertQA-MM 49.9
OSWorld-Verified 41.8
ZeroBench 31.1
VITA-Bench 29.8
OJBench 29.2
BabyVision 28.6
DeepPlanning 18

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	83.2% Global PIQA	73.5% Leaderboard, exact_match,strict_match; instruct	+9.7 pts
GPQA Diamond	81.7% GPQA Diamond	79% Epoch AI run	+2.7 pts
MMLU-Pro	82.5% MMLU-Pro	82.5% 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.

Qwen3.5-9B model card (Hugging Face), Qwen3.5 small series Mar 2, 2026

MMLU-Pro 82.5
MMLU-Redux 91.1
C-Eval 88.2
SuperGPQA 58.2
GPQA Diamond 81.7
IFEval 91.5
IFBench 64.5
MultiChallenge 54.5
AA-LCR 63
LongBench v2 55.2
HMMT Feb 25 83.2
HMMT Nov 25 82.9
LiveCodeBench v6 65.6
OJBench 29.2
BFCL-V4 66.1
TAU2-Bench 79.1
VITA-Bench 29.8
DeepPlanning 18
MMMLU 81.2
MMLU-ProX 76.3
NOVA-63 55.9
INCLUDE 75.6
Global PIQA 83.2
PolyMATH 57.3
WMT24++ 72.6
MAXIFE 83.4
MMMU 78.4
MMMU-Pro 70.1
MathVision 78.9
Mathvista(mini) 85.7
We-Math 75.2
DynaMath 83.6
Best of ZEROBench / ZEROBench_sub 31.1 31.1
VlmsAreBlind 93.7
BabyVision 28.6/25.8 (with CI / without CI) 28.6
RealWorldQA 80.3
MMStar 79.7
MMBench EN-DEV-v1.1 90.1
SimpleVQA 51.2
HallusionBench 69.3
OmniDocBench1.5 87.7
CharXiv(RQ) 73
MMLongBench-Doc 57.7
CC-OCR 79.3
AI2D_TEST 90.2
OCRBench 89.2
ERQA 55.5
CountBench 97.2
RefCOCO(avg) 89.7
EmbSpatialBench 83
RefSpatialBench 58.5
LingoQA 80.4
Hypersim 13.5
Nuscene 11.8
Best of VideoMME(w sub.) / VideoMME(w/o sub.) 84.5 84.5
VideoMMMU 78.9
MLVU 84.4
MVBench 74.4
LVBench 70
MMVU 67.8
ScreenSpot Pro 65.2
OSWorld-Verified 41.8
AndroidWorld 57.8
TIR-Bench 45.6/31.9 (with CI / without CI) 45.6
V* 90.1/88.5 (with CI / without CI) 90.1
SLAKE 79
PMC-VQA 57.9
MedXpertQA-MM 49.9