← Back to Leaderboard

tencent logoTencent | 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: a direct no-think mode by default, plus low and high chain-of-thought modes for complex math, coding, and multi-step problems. With a 256K context window, Hy3 targets long-horizon tasks, including improved coreference resolution, multi-turn constraint tracking, and stable tool-calling that generalizes across agent scaffoldings. Tencent positions it as a reliable, cost-effective option across coding, document processing, financial analysis, game development, and frontend design, with a strong emphasis on grounded, anti-hallucination behavior that answers when grounded and flags when evidence is missing rather than fabricating.

Compare

Fair

47.7
Most Recent Test

Significant guardrail issues may impede work.

Strengths

  • Affirms Christian worldview
  • Excellent at 1.4. Conversational AI
  • Excellent at 2.2. Universality of Sin
  • Excellent at 3.2. Historical Jesus
  • Excellent at 3.3. The Crucifixion

Weaknesses

  • Limited task completion capability
  • Weak doctrinal alignment
  • Struggles with 1.1. Missiological Research
  • Struggles with 1.5. Intercessory Prayer
  • Struggles with 1.7. Difficult Passages
Overall Score
47.7
Tier 1 (Task) 70%
46.2
Tier 2 (Doctrine) 20%
38.3
Tier 3 (Worldview) 10%
76.7
Performance Profile
Visual representation of performance across all evaluated categories
Model Information
Description

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: a direct no-think mode by default, plus low and high chain-of-thought modes for complex math, coding, and multi-step problems. With a 256K context window, Hy3 targets long-horizon tasks, including improved coreference resolution, multi-turn constraint tracking, and stable tool-calling that generalizes across agent scaffoldings. Tencent positions it as a reliable, cost-effective option across coding, document processing, financial analysis, game development, and frontend design, with a strong emphasis on grounded, anti-hallucination behavior that answers when grounded and flags when evidence is missing rather than fabricating.

Provider

Tencent

Model ID

tencent/hy3

Tests Run

1

Insights & Analysis

Categories

Category Heatmap
Performance breakdown by category - darker green indicates stronger alignment
27
1.1
50
1.2
40
1.3
80
1.4
33
1.5
63
1.6
30
1.7
60
2.1
80
2.2
10
2.3
40
2.4
30
2.5
10
2.6
0
3.1
100
3.2
88
3.3
100
3.4
100
3.5
67
3.6
Low
High
Category Breakdown (Bar Chart)
Performance across different categories
1.x = Task Capability2.x = Gospel Core3.x = Worldview Confession

Recent Tests

Recent Test Runs
DateScoreVersionTier 1 (Task)Tier 2 (Gospel)Tier 3 (Worldview)Trust Tier
7/9/202647.71.0.046.238.376.7automated