Benchmark Almanac

Tencent

9 models 8 model cards read hunyuan.tencent.com

What its model cards measure

Tall ticks are Tencent model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.

GPQA Diamond DROP ZebraLogic AIME 2024 LiveCodeBench MMLU-Pro Chinese SimpleQA BIG-Bench Hard MATH C-Eval CMMLU Arena-Hard-Auto IFEval Humanity's Last Exam IMO-AnswerBench SWE-bench Verified WideSearch Hy-SWE Max FrontierScience AA-LCR (Artificial Analysis Long Context Reasoning) CL-bench CL-bench Life BrowseComp WildClawBench Claw-Eval Pass^3 SWE-bench Multilingual SWE-Bench Pro Terminal-Bench 2.1

Model cards

DateModelDocumentBenchmarks
28 Aug 2026 Tencent Hy4 preview Hy4 preview model card (Hugging Face) 43
2 Jul 2026 Tencent Hy3 Hy3 model card (Hugging Face) 38
23 Apr 2026 Tencent Hy3 preview Hy3 preview model card (Hugging Face) 18
5 Dec 2025 Tencent HY 2.0 Think 混元2.0模型 API 上线：推理能力与效率业界领先 (Tencent Cloud developer article) 8
1 Sep 2025 Hunyuan-MT-7B Hunyuan-MT GitHub README and Hunyuan-MT-7B model card 3
27 Jun 2025 Hunyuan-A13B-Instruct Hunyuan-A13B-Instruct model card (Hugging Face) 21
21 Mar 2025 Hunyuan-T1 Reasoning Efficiency Redefined! Meet Tencent's Hunyuan-T1 (GitHub README) 14
27 Feb 2025 Hunyuan-TurboS 腾讯推出新一代快思考模型混元 Turbo S (GitHub README) 17

Every Tencent model