Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic workflows and production use. It supports configurable reasoning levels across disabled, low, and high modes, allowing it to adapt to various task requirements.
Author
T
TencentRelease Date
2026-04-22
Knowledge Cutoff
—
License
Proprietary
I/O Format
Context Length
262K / 262K
API I/O (1M)
—
How to Use
—
Output Speed
85 tok/sArena Overall
1419Intelligence Index
41.9Coding Index
36.5Math Index
—LiveBench
—ForecastBench
—GPQA Diamond
86.7%HLE
25.5%MMLU-Pro
—AIME 2025
—MATH-500
—LB Reasoning
—LB Math
—LB Data Analysis
—LiveCodeBench
—LB Coding
—LB Agentic
—TAU2
92.7%TerminalBench
34.1%SciCode
41.2%IFBench
63.1%AA-LCR
0.5Hallucination (HHEM)
—Factual Consistency (HHEM)
—LB Language
—LB Instruction Following
—1 / 3
Swipe to compare
Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic workflows and production use. It supports configurable reasoning levels across disabled, low, and high modes, allowing it to adapt to various task requirements.
Author
T
TencentRelease Date
2026-04-22
Knowledge Cutoff
—
License
Proprietary
I/O Format
Context Length
262K / 262K
API I/O (1M)
—
How to Use
—
Output Speed
85 tok/sArena Overall
1419Intelligence Index
41.9Coding Index
36.5Math Index
—LiveBench
—ForecastBench
—GPQA Diamond
86.7%HLE
25.5%MMLU-Pro
—AIME 2025
—MATH-500
—LB Reasoning
—LB Math
—LB Data Analysis
—LiveCodeBench
—LB Coding
—LB Agentic
—TAU2
92.7%TerminalBench
34.1%SciCode
41.2%IFBench
63.1%AA-LCR
0.5Hallucination (HHEM)
—Factual Consistency (HHEM)
—LB Language
—LB Instruction Following
—