Hunyuan-T1
Tencent·
LLMsreasoningcloud
Overview
Tencent's deep-thinking reasoning flagship, built on the TurboS Hybrid-Transformer-Mamba MoE base and post-trained with large-scale curriculum reinforcement learning. First Hunyuan entry in the frontier reasoning race.
Capabilities and innovations
Deep-thinking chain-of-thought reasoningHybrid-Transformer-Mamba MoE base (TurboS)Fast decoding on long inputsFirst ultra-large Mamba-hybrid reasoning modelCurriculum reinforcement-learning post-training
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 69.3% | vendor GPQA Diamond 69.3, vendor benchmark table at release. |
| MMLU-Pro | 87.2% | vendor MMLU-Pro 87.2, vendor benchmark table at release (second behind o1 at the time). |
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: benchmark did not exist at release
- MMLU: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
More from Tencent
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.