ModelBench
Tencent logo
Models / Tencent

Tencent: Hy3

tencent/hy3

Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort:...

Input modalitiesT
Output modalitiesT
Source-linkedClosed weights
Report data
Context window262.144K
Maximum outputNot documented
Provider offers0
Knowledge cutoffUndisclosed
Capabilities

What this model supports

ReasoningNo
Tool callingNo
Structured outputNo
AttachmentsNo
Temperature controlNo
Open weightsNo
Creator
Tencent
Family
Not documented
Modalities
Text
License
Not documented
Release date
Not documented
Model ID
tencent/hy3
Market availability

Provider pricing

Report price →
No known providers available.
Evidence

Published benchmark results

Design Arena: 3d vs top 8 models
Z.ai: GLM 5.2
1371.0
Claude Fable 5
1366.0
Anthropic: Claude Opus 4.6
1349.0
Kimi K2.6
1348.0
xAI: Grok 4.5
1338.0
Z.ai: GLM 5.1
1336.0
DeepSeek V4 Pro
1326.0
Tencent: Hy3
1187.0
Design Arena: codecategories vs top 8 models
Z.ai: GLM 5.2
1351.0
Claude Fable 5
1345.0
Anthropic: Claude Opus 4.6
1336.0
Anthropic: Claude Opus 4.7
1330.0
xAI: Grok 4.5
1327.0
Z.ai: GLM 5.1
1321.0
Kimi K2.6
1320.0
Tencent: Hy3
1191.0
Design Arena: dataviz vs top 8 models
Z.ai: GLM 5.1
1366.0
Claude Fable 5
1363.0
xAI: Grok 4.5
1330.0
Anthropic: Claude Opus 4.6
1322.0
Anthropic: Claude Opus 4.7
1322.0
Anthropic: Claude Sonnet 4.6
1316.0
Z.ai: GLM 5.2
1316.0
Tencent: Hy3
1179.0
Design Arena: gamedev vs top 8 models
Claude Fable 5
1383.0
Z.ai: GLM 5.2
1353.0
Claude Sonnet 5
1345.0
GPT-5.5
1344.0
Anthropic: Claude Opus 4.6
1339.0
Anthropic: Claude Opus 4.7
1336.0
MiMo-V2.5-Pro
1334.0
Tencent: Hy3
1194.0
Design Arena: website vs top 8 models
Z.ai: GLM 5.2
1346.0
Claude Fable 5
1338.0
Anthropic: Claude Opus 4.6
1331.0
Anthropic: Claude Opus 4.7
1329.0
xAI: Grok 4.5
1325.0
Anthropic: Claude Sonnet 4.6
1320.0
Claude Sonnet 5
1314.0
Tencent: Hy3
1189.0