7 points | by wrpark 6 hours ago ago
1 comments
By their model cards, Tencent's new Hunyuan Instruct models supports: - Hybrid Reasoning Support - Ultra-Long(256k) Context Understanding - Enhanced Agent Capabilities - Efficient Inference(GQA)
And the Hunyuan 7B model showed better scores than the Qwen3 8B model in the DROP, AIME 2024, and AIME 2025 benchmarks according to this thread.: https://www.threads.com/@sung.kim.mw/post/DM7JPKbRL_E
By their model cards, Tencent's new Hunyuan Instruct models supports: - Hybrid Reasoning Support - Ultra-Long(256k) Context Understanding - Enhanced Agent Capabilities - Efficient Inference(GQA)
And the Hunyuan 7B model showed better scores than the Qwen3 8B model in the DROP, AIME 2024, and AIME 2025 benchmarks according to this thread.: https://www.threads.com/@sung.kim.mw/post/DM7JPKbRL_E