Tencent Hy4
Tencent Hy4 places in the top three on 2 benchmarks.
35
Benchmarks with a score
0
Ranked first, out of five or more models
2
In the top three
Tencent Nov 2026 Tencent Hy4
Closed weights Flagship
No Epoch Capabilities Index score yet
- Terminal-Bench 4.0 39.1% #15 /40
- GDP.pdf 43.6% #12 /19
- OSWorld 2.1 78.9% #4 /7
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 | |
|---|---|---|---|---|
| BullshitBench V2 | 57% | #42 of 129 | Official leaderboard | |
| Extended NYT Connections | 83.5% | #38 of 103 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Terminal-Bench-Science | 19.8% | #11 of 12 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OpenDesign Arena | 90.0 | #4 of 21 | Official leaderboard | |
| DAXBench v2 | 85.7% | #11 of 47 | Official leaderboard | |
| DesignArena Game Dev | 1,269 | #34 of 143 | Official leaderboard | |
| DesignArena 3D Design | 1,258 | #33 of 136 | Official leaderboard | |
| Vibe Code Bench | 81% | #32 of 120 | Official leaderboard | |
| FrontierCode 2.0 | 31% | #2 of 6 | Lab model card | |
| DesignArena Code | 1,256 | #48 of 143 | Official leaderboard | |
| DesignArena Website | 1,253 | #53 of 151 | Official leaderboard | |
| ProgramBench almost solved | 30% | #4 of 8 | Lab model card | |
| DesignArena UI Component | 1,194 | #74 of 139 | Official leaderboard | |
| PostTrainBench 1.2 | 35.2% | #2 of 3 | Lab model card | |
| DesignArena Image to HTML | 1,137 | #45 of 51 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AutomationBench | 56.2% | #5 of 47 | Lab model card | |
| Agent Arena | 5.26% | #14 of 56 | Official leaderboard | |
| Terminal-Bench 4.0 | 39.1% | #15 of 40 | Official leaderboard | |
| Terminal-Bench 4.1 | 37.9% | #18 of 43 | Official leaderboard | |
| OSWorld 2.1 | 78.9% | #4 of 7 | Lab model card | |
| DuelLab GameBench 2 | 0.00 | #57 of 57 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FoodTruck Bench | $67,598 | #3 of 59 | Official leaderboard | |
| APEX-Agents 1.1 | 85.1% | #4 of 65 | Official leaderboard | |
| Tax Agent Bench | 34.4% | #18 of 84 | Official leaderboard | |
| Vals Finance Agent v2.1 | 64.7% | #14 of 40 | Official leaderboard | |
| Legal Research Bench | 41.4% | #34 of 93 | Official leaderboard | |
| Vals Finance Index | 58% | #16 of 40 | Official leaderboard | |
| Agents' Last Exam v1.1 | 41% | #6 of 14 | Official leaderboard | |
| Vals Index v2.2 | 59.4% | #16 of 37 | Official leaderboard | |
| Vals Legal Index | 70.5% | #17 of 38 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 12.5% | #18 of 38 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 66.2% | #32 of 42 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GDP.pdf | 43.6% | #12 of 19 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 94.0 | #32 of 185 | Official leaderboard | |
| CWE-bench 2 | 64.1% | #7 of 14 | Official leaderboard |
Its lab's numbers next to 2 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 |
|---|---|---|---|
| Terminal-Bench 4.0 | 44.4% Terminal Bench 4.0 | 39.1% Leaderboard, max, Claude Code | +5.3 pts |
| Agents' Last Exam v1.1 | 41% Agents' Last Exam v1.1* | 41% Leaderboard, high, Claude Code | Same |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.