Back to Models
Tencent: Hy3
Tencenttext9.0 / 10 Overall Rating
Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent featuring 21B active parameters and top-8 routing. Built for reasoning and agentic workloads, it offers a 262K context window and an exceptionally high 128K max output limit at a fraction of baseline API costs.
Context Window
262K
Knowledge Cutoff
2025
Max Output
16K
License
Proprietary
Where it Excels (Pros)
- Massive 262K token context window
- Significantly cheaper than GPT-4o
- Large 128K max output limit
Limitations (Cons)
- Unproven benchmarks vs Claude 3.5
- Limited ecosystem and tool integration
Benchmark Breakdown
Reasoning (MMLU)82%
Coding (HumanEval)78%
Mathematics (MATH)76%
Long-Context Retrieval95%
Pricing Matrix
Input cost / 1M tokens$0.13
Output cost / 1M tokens$0.53
Compare Tencent: Hy3 with other frontier models: