Benchmark Almanac

Qwen3.5 397B-A17B

Qwen3.5 397B-A17B ranks first on 7 of the 161 benchmarks with five or more models scored.

Qwen Released February 13th, 2026 Family Qwen3.5 The lab's model card Compare with others
163
Benchmarks with a score
7
Ranked first, out of five or more models
28
In the top three
Qwen Feb 2026

Qwen3.5 397B-A17B

Open weights Flagship

147.9 #77 of 256
Epoch Capabilities Index
Self-Reported Index 149.8 #96 of 404
  • MMMU-Pro 79% #20 /90
  • IFBench 76.5% #18 /80
  • AA-LCR (Artificial Analysis Long Context Reasoning) 68.7% #11 /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

BenchmarkScoreRankSource
MMLU-Pro 87.8% #6 of 167 Lab model card
MMLU-Redux 94.9% #2 of 45 Lab model card
C-Eval 93% #4 of 46 Lab model card
NOVA-63 59.1% #1 of 11 Lab model card
HallusionBench 71.4% #3 of 31 Lab model card
BullshitBench V2 78% #14 of 129 Official leaderboard
BullshitBench 65.5% #12 of 107 Official leaderboard
SEA-HELM 76.2 #8 of 71 Official leaderboard
Judgemark v2.1 85.5 #5 of 42 Official leaderboard
IFEval 92.6% #17 of 135 Lab model card
MMLU-ProX 84.7% #3 of 23 Lab model card
SuperGPQA 70.4% #6 of 35 Lab model card
KINA 43% #7 of 37 Official leaderboard
MultiChallenge 67.6% #8 of 42 Lab model card
INCLUDE 85.6% #6 of 31 Lab model card
LongBench v2 63.2% #7 of 35 Lab model card
IFBench 76.5% #18 of 80 Lab model card
LMArena (Elo) 1,441 #76 of 335 Official leaderboard
MMMLU 88.5% #18 of 71 Lab model card
MAXIFE 88.2% #3 of 11 Lab model card
HELM Arabic 80.2 #8 of 28 Official leaderboard
AA-LCR (Artificial Analysis Long Context Reasoning) 68.7% #11 of 38 Lab model card
Humanity's Last Exam 48.3% #47 of 154 Lab model card
EQ-Bench 3 1,451 #28 of 78 Official leaderboard
Longform Creative Writing (EQ-Bench) 68.3 #46 of 119 Official leaderboard
Global PIQA 89.8% #6 of 14 Lab model card
SimpleVQA 67.1% #13 of 30 Lab model card
WideSearch 74% #9 of 19 Lab model card
HELM Arabic Enterprise 65.3 #12 of 23 Official leaderboard
Creative Writing v3 66.3 #64 of 120 Official leaderboard
Judgemark v4 54.2 #23 of 40 Official leaderboard
Extended NYT Connections 58.9% #67 of 103 Official leaderboard
Seal-0 46.9% #5 of 7 Lab model card
WMT24++ 78.9% #15 of 15 Lab model card

Math and science

BenchmarkScoreRankSource
PolyMath 73.3% #4 of 24 Lab model card
HMMT 2025 94.8% #13 of 54 Lab model card
General365 57.7% #6 of 23 Official leaderboard
GPQA Diamond 86.4% #74 of 229 Epoch AI run
OTIS Mock AIME 2024-2025 88.9% #72 of 212 Epoch AI run
DTBench 87.5% #64 of 180 Official leaderboard
HMMT Nov 2025 92.7% #11 of 28 Lab model card
HMMT 2026 87.9% #17 of 31 Official leaderboard
MathArena Apex Shortlist 60.1% #22 of 38 Official leaderboard
IMO-AnswerBench 80.9% #23 of 37 Lab model card
Echo General AI Prediction Leaderboard 986 #12 of 17 Official leaderboard
AIME 2026 94.2% #22 of 31 Official leaderboard
FrontierMath Tiers 1-3 v2 31.2% #67 of 86 Official leaderboard
ResearchClawBench 14.2 #18 of 23 Official leaderboard
USAMO 2026 36.3% #8 of 9 Official leaderboard
FrontierChallenge 4.10% #10 of 11 Official leaderboard

Coding

BenchmarkScoreRankSource
SWE-bench Verified 76.4% #39 of 155 Lab model card
LiveCodeBench v6 83.6% #23 of 90 Lab model card
DesignArena Website 1,184 #84 of 151 Official leaderboard
DesignArena 3D Design 1,144 #77 of 136 Official leaderboard
DesignArena Code 1,170 #82 of 143 Official leaderboard
SWE-bench Multilingual 69.3% #45 of 73 Lab model card
DesignArena SVG 1,130 #65 of 103 Official leaderboard
DesignArena UI Component 1,154 #88 of 139 Official leaderboard
DesignArena Game Dev 1,130 #91 of 143 Official leaderboard
DesignArena Data Visualization 1,173 #76 of 118 Official leaderboard
Code Arena 1,400 #85 of 128 Official leaderboard
AgenticDataBench 47.1% #2 of 3 Official leaderboard
EvoCode-Bench 4.60% #11 of 13 Official leaderboard
ALE-Bench 1,007 #108 of 126 Official leaderboard
Vision2Web S2-2026 19.2% #9 of 9 Official leaderboard

Agents and tools

BenchmarkScoreRankSource
VITA-Bench 49.7% #1 of 16 Lab model card
BFCL V4 72.9% #3 of 30 Lab model card
EnvSimBench 42.3% #1 of 7 Official leaderboard
tau3-bench 87.9% #1 of 7 Official leaderboard
ClawProBench 64.2% #12 of 49 Official leaderboard
DeepPlanning 34.3% #3 of 10 Lab model card
tau2-bench 86.7% #24 of 79 Lab model card
AutoResearchBench 6.97% #4 of 13 Official leaderboard
BrowseComp-ZH 70.3% #6 of 19 Lab model card
LisanBench 1293.3 #51 of 133 Official leaderboard
BrowseComp 78.6% #32 of 81 Lab model card
Toolathlon 38.3% #23 of 50 Official leaderboard
AndroidWorld 66.8% #12 of 23 Lab model card
Terminal-Bench 2.0 52.5% #34 of 63 Lab model card
SnakeBench 23 #110 of 188 Official leaderboard
ScreenSpot-Pro 65.6% #19 of 32 Lab model card
Mystery Game Puzzles 18% #55 of 84 Epoch AI run
Claw-Eval Pass^3 56.8% #16 of 24 Official leaderboard
Clocktower Radio 1,478 #16 of 23 Official leaderboard
Claw-Eval 74.2% #17 of 24 Official leaderboard
MobileWorld 36.8% #13 of 18 Official leaderboard
OSWorld-Verified 62.2% #30 of 41 Lab model card
DeepTerminalBench 24% #7 of 9 Official leaderboard
GameHorizon 62.3 #29 of 37 Official leaderboard
MCPMark 46.1% #8 of 10 Lab model card
LiveClawBench 57% #14 of 17 Official leaderboard
Pencil Puzzle Bench 0% #47 of 53 Official leaderboard
Toolathlon-Verified 40.7% #23 of 25 Official leaderboard
τ³-bench Banking 9.79% #25 of 25 Official leaderboard
Claw-Eval-Live 49.5% #12 of 12 Official leaderboard
WeaveBench 0.90% #9 of 9 Official leaderboard

Professional work

BenchmarkScoreRankSource
PMC-VQA 64.2% #2 of 10 Lab model card
MedXpertQA-MM 70% #5 of 21 Lab model card
SLAKE 79.9% #5 of 10 Lab model card
ProfBench 45.1% #44 of 72 Official leaderboard
SpreadsheetBench 2 11.2% #7 of 8 Official leaderboard
APEX-Agents 24.9% #58 of 61 Collected by Epoch AI, source not named
BankerToolBench v1.1 2.70% #10 of 10 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
MathVista 90.3% #2 of 70 Lab model card
MMMU 85% #4 of 103 Lab model card
MMBench-EN v1.1 93.7% #1 of 23 Lab model card
MMStar 83.8% #2 of 36 Lab model card
OCRBench 93.1% #2 of 34 Lab model card
RefSpatial-Bench 73.6% #1 of 16 Lab model card
Nuscene 16.0 #1 of 8 Lab model card
DynaMath 86.3% #3 of 23 Lab model card
MVBench 77.6% #3 of 23 Lab model card
We-Math 87.9% #2 of 13 Lab model card
VideoMME v2 (with subtitles) 39.1% #5 of 31 Official leaderboard
MDPBench 83.6 #2 of 12 Official leaderboard
AI2D 93.9% #9 of 52 Lab model card
MMLongBench-Doc 61.5% #3 of 17 Official leaderboard
V* 95.8% #3 of 17 Lab model card
CountBench 97.2% #5 of 28 Lab model card
EmbSpatialBench 84.5% #3 of 16 Lab model card
OVO-S-Bench 52.1% #3 of 15 Official leaderboard
RefCOCO / RefCOCO+ / RefCOCOg 92.3% #6 of 29 Lab model card
CC-OCR 82% #3 of 14 Lab model card
Video-MME 87.5% #11 of 51 Lab model card
MMMU-Pro 79% #20 of 90 Lab model card
SpatialWorld Physical 12.2% #2 of 9 Official leaderboard
SUNRGBD 38.3 #2 of 9 Lab model card
MLVU 86.7% #5 of 22 Lab model card
MATH-Vision (MathVision) 88.6% #12 of 51 Lab model card
Video-MMMU 84.7% #9 of 36 Lab model card
MMVU 75.4% #7 of 26 Lab model card
ERQA 67.5% #10 of 36 Lab model card
GeoGuesser environment (FineEnvs) 0.34 #5 of 14 Official leaderboard
Vision Arena 1,246 #49 of 136 Official leaderboard
RealWorldQA 83.9% #14 of 38 Lab model card
ODinW-13 47% #5 of 13 Lab model card
LingoQA 81.6% #4 of 10 Lab model card
LVBench 75.5% #13 of 31 Lab model card
CharXiv 80.8% #32 of 75 Lab model card
AgentWorldBench 54.7 #6 of 13 Official leaderboard
BabyVision 43.3% #15 of 27 Official leaderboard
Hypersim 12.5 #5 of 9 Lab model card
SpatialWorld Digital 26% #5 of 9 Official leaderboard
CiteVQA 18.3% #9 of 16 Official leaderboard
PerceptionBench (Moonshot) 47.5% #9 of 16 Official leaderboard
Claw-Eval-MM 50.2% #7 of 11 Official leaderboard
MineBench 1,275 #51 of 75 Official leaderboard

Safety and security

BenchmarkScoreRankSource
LiveSecBench 81.5 #5 of 34 Official leaderboard
BioTIER Permit 99.8% #14 of 69 Official leaderboard
Molecular Biology Capabilities Test 52.5% #29 of 90 Official leaderboard
nocot-bench (NCRI) 80.0 #60 of 185 Official leaderboard
BioTIER Refuse 35.7% #39 of 69 Official leaderboard
ABC Bench Screening Evasion 56.7% #10 of 17 Official leaderboard
Human Pathogen Capabilities Test 48.3% #52 of 86 Official leaderboard
SMDD-Bench 27.5% #5 of 7 Official leaderboard
World-Class Bio 34.2% #54 of 74 Official leaderboard
ABC Bench Liquid-Handling Robot 57.3% #26 of 34 Official leaderboard
ABC Bench Fragment Design 76.3% #22 of 27 Official leaderboard
SecCodeBench 68.3% #1 of 1 Lab model card

Other

BenchmarkScoreRankSource
LMCA 37.9% #65 of 147 Official leaderboard
Conceptual Reasoning Index 49.7 #70 of 151 Official leaderboard
SAGE v1.1 (Vals) 85.5% #15 of 29 Official leaderboard
ACCoRD 55.6 #104 of 152 Official leaderboard

Its lab's numbers put it at 149.8, 1.9 points above its ECI

The Self-Reported Index fits one score to the 71 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 148.9 to 150.7. How we compute it.

Its lab's numbers next to 7 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
BabyVision 52.3% BabyVision (52.3/43.3) 43.3% Leaderboard +9.0 pts
AIME 2026 91.3% AIME26 94.2% Leaderboard −2.9 pts
GPQA Diamond 88.4% GPQA 86.4% Epoch AI run, none +2.0 pts
MATH-Vision (MathVision) 88.6% MathVision 88.6% Leaderboard Same
MMLongBench-Doc 61.5% MMLongBench-Doc 61.5% Leaderboard, VL; 17B activated (397B total) Same
MMLU-Pro 87.8% MMLU-Pro 87.8% Leaderboard Same
Toolathlon 38.3% Tool Decathlon 38.3% Leaderboard Same

Qwen's models over time

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

50 100 150 2023 2024 2025 2026 2027 Qwen3.8 Max (0902) Qwen3.7-Plus Qwen3.5-9B Qwen3-14B Qwen2.5-72B Qwen2.5-Coder (1.5B) Qwen2.5-Coder-32B 50 100 150 2023 2024 2025 2026 2027 Qwen3.8 Max (0902) Qwen 3.6 Flash

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.

Qwen3.5-397B-A17B model card Feb 16, 2026