The AI Loadout
←Back to models
Checked Active: 2026-11-02
Meituan

LongCat-2.5

Low-cost long-horizon coding agents and visual coding, such as building web pages from screenshots or designs.

Evaluation Analysis: Released Nov 1, 2026 by Meituan's LongCat team. A native multimodal mixture-of-experts model with 1.6 trillion total parameters and about 48 billion active per token. Meituan reports 88.4 on Terminal Bench v2.1, 38.5 on Terminal Bench v4.0 and 64.6 on DeepSWE v1.1, its own runs, which put it between Claude Opus 4.8 and Claude Opus 5 on most of its charts. Cached input $0.006 per 1M tokens. Thinking can be turned on or off. OpenAI-compatible Chat Completions and Responses endpoints plus an Anthropic Messages endpoint (model LongCat-2.5), so Claude Code and Codex can point at it. Prepaid balance; Meituan says unused balance can be refunded. Prices are a limited-time discount.
Context Window1,048,576 tokens (262,144 max output)
Input Price / 1M$0.30 (limited-time discount)
Output Price / 1M$1.20 (limited-time discount)
Release Date2026-11-01

Specs and prices checked 2026-11-02 against the official page.

Used in these setups

Related skills and personas

Agent skills and bot personas that pair well with LongCat-2.5.

Browse full directories:All Skills (>35)·All Personas (>15)

Want tested skills and personas installed in one command? See the AI Loadout packs, from $9.