Monday Jul 6
STARTUPSOther

Meituan's Stealth AI Tops OpenRouter

6JUL
OWL ALPHA1.6T PARAMS

A mystery AI model was quietly winning for two months. Meituan revealed Owl Alpha was its LongCat-2, trained with zero Nvidia chips. It already leads coding-agent use on OpenRouter, built entirely on Chinese chips.

The mystery lasted nine weeks before anyone knew who built it. Real coders had already made their pick.

The model carries 1.6 trillion parameters, with 48 billion active at once. Every training token, more than 35 trillion of them, ran on domestically made chips instead of imported silicon.

Two named analysts weighed in fast. One said export limits just pushed the shift to home-grown hardware. The other said it quiets doubts about a big domestic chip cluster.

full brief & sources

⚡ Why this matters

  • First trillion-parameter-class model trained end to end with zero Nvidia chips, while still winning real developer usage.
  • It won on merit before anyone knew whose model it was. No brand halo, no hype push.
  • Undercuts the idea that US export controls freeze Chinese AI progress.

🔍 What happened

  • Since late April, an anonymous model called Owl Alpha ran on OpenRouter with no listed developer.
  • On June 29, Meituan's LongCat account confirmed: Owl Alpha on OpenRouter, that's us.
  • The model is LongCat-2: 1.6 trillion total parameters, 48 billion active.
  • Trained on over 50,000 domestic Chinese AI chips, more than 35 trillion tokens, no Nvidia hardware at any stage.
  • Leads Hermes agent usage, ranks second in Claude Code and third in OpenClaw usage, per Meituan's own benchmarks.
  • Now open-sourced, with an MIT license and no regional usage restrictions.

💬 Smart takes

  • Yuchen Jin, AI analyst: export controls on Nvidia chips will just accelerate development of AI that runs on Chinese chips.
  • TP Huang, tech analyst: the launch quiets doubts about Huawei's Atlas-950 chip clusters.
  • Skeptic: Meituan's own benchmark numbers aren't independently verified, and the stealth launch makes early praise hard to separate from marketing.

🧭 Where this goes

  1. Likelymore Chinese labs copy the stealth-launch-then-reveal playbook.
  2. LikelyOpenRouter usage rankings become a real signal, not just a hype gauge.
  3. PossibleWestern developers adopt LongCat-2 for cost reasons, regardless of its origin.
  4. Wild Cardthis reshapes the export-control debate in Washington within the year.

🥄 The Spoon Take

Meituan didn't announce a model and hope people cared. It let the model earn its reputation anonymously, then took credit. That's a new playbook, and it worked precisely because export controls couldn't stop it.

🤔 Pushback

Self-reported benchmarks from the model's own maker aren't the same as independent, adversarial evaluation.

·Meituan LongCat - Reveal post·VentureBeat - LongCat-2 open source·The Decoder - No-Nvidia training