Benchmark Almanac

Tencent Hy3

Tencent Hy3 ranks first on 2 of the 67 benchmarks with five or more models scored.

Tencent Released July 2nd, 2026 Family Hy3 The lab's model card Compare with others
78
Benchmarks with a score
2
Ranked first, out of five or more models
3
In the top three
Tencent Jul 2026

Tencent Hy3

Open weights Flagship

152.0 #84 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • AA-LCR (Artificial Analysis Long Context Reasoning) 73.4% #4 /38
  • NL2Repo-Bench 45.6% #18 /30
  • Toolathlon-Verified 56.8% #16 /25

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
AA-LCR (Artificial Analysis Long Context Reasoning) 73.4% #4 of 38 Lab model card
CL-bench 23.8% #1 of 6 Lab model card
LMArena (Elo) 1,454 #59 of 335 Official leaderboard
Humanity's Last Exam 53.2% #31 of 154 Lab model card
WideSearch 76.4% #5 of 19 Lab model card
StudyArena Text 1,238 #35 of 85 Official leaderboard
CL-bench Life 17% #1 of 2 Lab model card
DesignArena ASCII Art 1,153 #43 of 77 Official leaderboard
Continual Learning Exam 27.3% #8 of 11 Official leaderboard
Extended NYT Connections 41.2% #75 of 103 Official leaderboard
Hemingway-bench 971 #36 of 42 Official leaderboard

Math and science

BenchmarkScoreRankSource
GPQA Diamond 90.4% #32 of 302 Lab model card
IMO-AnswerBench 90% #4 of 37 Lab model card
PHYBench 77.4% #1 of 5 Lab model card
DTBench 89.3% #55 of 180 Official leaderboard
MathArena (Apex, IMO, USAMO) 38.7% #6 of 12 Lab model card
HorizonMath 7.10% #2 of 3 Lab model card
xbench-ScienceQA 66.6% #24 of 29 Official leaderboard
USAMO 2026 72% #5 of 6 Lab model card
Prophet Arena 84.5 #24 of 24 Official leaderboard
PaperBenchX 2.15% #10 of 10 Official leaderboard
ArXivMath 52.2% #9 of 9 Lab model card
FrontierScience 48.1% #8 of 8 Lab model card
CMT-Benchmark 37.8% #2 of 2 Lab model card
Hy-Euler pro 24.2% #1 of 1 Lab model card
Hy-Math 60.9% #1 of 1 Lab model card

Coding

BenchmarkScoreRankSource
SWE-bench Verified 78% #28 of 155 Lab model card
StructEval 76.9% #8 of 28 Official leaderboard
DesignArena SVG 1,194 #40 of 103 Official leaderboard
DesignArena 3D Design 1,212 #55 of 136 Official leaderboard
SWE-bench Multilingual 75.8% #31 of 73 Lab model card
SWE-Bench Pro 57.9% #35 of 81 Lab model card
ALE-Bench 1,516 #55 of 126 Official leaderboard
DesignArena Game Dev 1,198 #67 of 143 Official leaderboard
Code Arena 1,506 #60 of 128 Official leaderboard
DesignArena Website 1,222 #71 of 151 Official leaderboard
DesignArena Code 1,216 #71 of 143 Official leaderboard
DesignArena UI Component 1,195 #73 of 139 Official leaderboard
NL2Repo-Bench 45.6% #18 of 30 Lab model card
Hy-SWE Max 49% #2 of 3 Lab model card
DesignArena Data Visualization 1,163 #84 of 118 Official leaderboard
Tencent WorkBuddy Bench 70.9% #13 of 14 Official leaderboard
DeepSWE 28% #55 of 56 Lab model card
Doc2Harness 27.4% #19 of 19 Official leaderboard
Hy-Backend 2.0 25% #2 of 2 Lab model card
Hy-CompanyBench 41.7% #1 of 1 Lab model card

Agents and tools

BenchmarkScoreRankSource
Claw-Eval Pass^3 68.5% #2 of 13 Lab model card
SkillsBench 55.9% #5 of 22 Official leaderboard
BrowseComp 84.2% #21 of 81 Lab model card
DeepSearchQA 91% #7 of 23 Lab model card
MCP Atlas 79.1% #16 of 46 Lab model card
WildClawBench 49.7% #13 of 31 Official leaderboard
Toolathlon 48.5% #15 of 29 Lab model card
Toolathlon-Verified 56.8% #16 of 25 Official leaderboard
Terminal-Bench 2.1 71.7% #36 of 56 Lab model card
BabaIsBench 54% #26 of 32 Official leaderboard
Agent Arena -7.56% #46 of 56 Official leaderboard
TraceTrial-Bench 71.4% #8 of 9 Official leaderboard
Hy-SkillsWorld 45.8% #1 of 1 Lab model card
ProdBench (Tencent) 23% #1 of 1 Lab model card

Professional work

BenchmarkScoreRankSource
FoodTruck Bench $28,055 #23 of 59 Official leaderboard
CorpBench Work 78.6 #10 of 17 Official leaderboard
EnterpriseBench: CoreCraft 41% #34 of 51 Official leaderboard
DAYJOB: Healthcare 1.20% #27 of 37 Official leaderboard
DAYJOB: Finance 0.50% #32 of 39 Official leaderboard
APEX-Agents 25.6% #10 of 12 Lab model card
E-Bench-Code (Tencent) 64.4% #7 of 8 Official leaderboard
Tuesday Work Index 30.2 #39 of 44 Official leaderboard
YC-Bench $124,880 #54 of 59 Official leaderboard
$OneMillion-Bench 13% #13 of 13 Official leaderboard
OR-License 45.2% #11 of 11 Official leaderboard
E-Bench (Tencent) 48.5% #8 of 8 Official leaderboard
Hy-FinModelBench 69% #1 of 1 Lab model card

Safety and security

BenchmarkScoreRankSource
nocot-bench (NCRI) 72.5 #86 of 185 Official leaderboard
SUPERChem 54.9% #4 of 5 Lab model card

Other

BenchmarkScoreRankSource
Conceptual Reasoning Index 52.2 #62 of 151 Official leaderboard
ACCoRD 66.4 #66 of 152 Official leaderboard
LMCA 37.5% #69 of 147 Official leaderboard

Its lab's numbers put it at 152.0, number 84 of 404

The Self-Reported Index fits one score to the 21 capability results Tencent 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 149.4 to 154.7. How we compute it.

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
WildClawBench 53.6% WildClawBench (35, text-only) 49.7% Leaderboard, OpenClaw +3.9 pts
E-Bench (Tencent) 50.2% e-bench (internal) 48.5% Leaderboard, highest available reasoning +1.7 pts
SkillsBench 55.3% SkillsBench (79, text-only) 55.9% Leaderboard, Claude Code −0.6 pts

Its launch post

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

Hy3 model card (Hugging Face) Jul 2, 2026