Meituan Releases LongCat-2.5, a 1.6T-Parameter Agentic MoE Model

Meituan's LongCat team launched LongCat-2.5, a natively multimodal 1.6T-parameter MoE model with about 48B activated parameters, built for long-horizon agent and real-world software engineering tasks. It reportedly beats Claude-Opus-5 on agentic benchmarks and charges only $0.30 per million input tokens, though some suspect its architecture mirrors DeepSeek.

2026-11-02 ~ 2026-11-02 · 3 related posts

• Meituan's LongCat-2.5 lands as 1.6T MoE agent model, with DeepSeek copycat speculation — teortaxesTex · 2026-11-02
• Meituan releases LongCat-2.5: 1.6T-parameter native multimodal model beating Claude on agent benchmarks — 智东西 · 2026-11-02
• Meituan's LongCat-2.5 debuts: 1.6T-param multimodal agentic model at $0.30/1M input tokens — iamfakhrealam · 2026-11-02