Tencent Hy3

Composite indexes

• Artificial Analysis Intelligence Index#51 of 142
  25independentv4.3.2Artificial Analysis: Hy3 ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 4.0#53 of 88
  0.5%independentAAArtificial Analysis: Hy3 ↗

  Leader: Claude Mythos 5.1 at 60.9%

• Terminal-Bench 2.x#43 of 103
  71.7%vendorTerminal-Bench 2.1Hugging Face: tencent/Hy3 ↗
  64.4%independentAA, Terminal-Bench 2.1Artificial Analysis: Hy3 ↗

  Leader: Claude Fable 5.1 at 91.4%

• SciCode#45 of 94
  48.6%independentAAArtificial Analysis: Hy3 ↗

  Leader: Claude Opus 5.5 at 66.9%

• SWE-bench Verified#4 of 16
  78%vendorHugging Face: tencent/Hy3 ↗

  Leader: Claude Sonnet 5 at 85.2%

• SWE-bench Pro#16 of 31
  57.9%vendorHugging Face: tencent/Hy3 ↗

  Leader: Claude Opus 5.5 at 89.9%

Agents and tool use

• GDPval-AA#49 of 103
  1045independentGDPval-AA v2.1Artificial Analysis: Hy3 ↗

  Leader: Claude Opus 5.5 at 1846

Reasoning and knowledge

• Humanity’s Last Exam#52 of 147
  33.5%independentAA, no toolsArtificial Analysis: Hy3 ↗

  Leader: Claude Opus 5.5 at 61.4%

• GPQA Diamond#41 of 142
  90.4%vendorHugging Face: tencent/Hy3 ↗
  89.7%independentAAArtificial Analysis: Hy3 ↗

  Leader: GPT-6 Astra at 96.1%

• CritPt#50 of 139
  4.9%independentAAArtificial Analysis: Hy3 ↗

  Leader: GPT-5.6 Sol at 32.3%

Vision and long context

• AA Long Context Reasoning#44 of 139
  79%independentAAArtificial Analysis: Hy3 ↗

  Leader: Kimi K3 at 88.7%

Human preference

• LMArena Text#30 of 54
  1456independentarena entry "hy3", rank 60, 8047 votesLMArena Text leaderboard (default view, fetched 2026-09-25) ↗

  Leader: Claude Fable 5 at 1506

• LMArena WebDev#37 of 54
  1509independentarena entry "hy3", rank 46, 4821 votesLMArena WebDev leaderboard (fetched 2026-09-25) ↗

  Leader: Claude Opus 5.5 at 1818

Other published results