Meituan's LongCat-2.5 debuts: 1.6T-param multimodal agentic model at $0.30/1M input tokens
iamfakhrealam · x · 2026-11-02
Meituan's LongCat team released LongCat-2.5, a natively multimodal agentic model with 1.6T total parameters and 48B active, built for long-horizon agentic tasks.
Key points:
- Trained on very long tasks (up to 1M tokens), so it holds context across full projects
- Stronger visual understanding: can turn screenshots or designs into working code, plus creative design and 3D visual coding
- Learns from its own mistakes: real failed attempts are converted into new practice tasks
- Aggressive pricing: $0.30/1M uncached input, $0.006/1M cached input, $1.20/1M output tokens
- 10M free tokens for new signups, plus another 10M for existing users
Related event: Meituan Releases LongCat-2.5, a 1.6T-Parameter Agentic MoE Model(3 posts)→
More from coding & agent
- Vibe coding's 5-year leap: from a GPT-3 React button demo to an Opus 5.5 three.js world — bennash · 2026-11-03
- Stripe's Link Agent Wallet lands in Merge Agent Handler: agents pay without seeing card details — shensi · 2026-11-03
- Ask Your Agent to Reverse-Engineer HTTP Calls Into a CLI, Skip Slow Computer-Use — vivekhaldar · 2026-11-03
- The 'Illustrated LLM Book' repo is free: 12 chapters of runnable Colab notebooks — mdancho84 · 2026-11-03
- Google open-sources Elephant-Goldfish Model: agent skills for software design and review — adnan_hashmi · 2026-11-03
- CircleCI report: work in progress is the new software delivery bottleneck — rseroter · 2026-11-03