A newer version of this model is available: meituan-longcat/LongCat-2.0

A cloaked owl raises her head. Owl Alpha. Revealed: LongCat-2.0 by Meituan. You have my attention.

Owl Alpha was Meituan's LongCat-2.0: native tool use, a 1M-token context, and open weights under MIT.

By Stealth Models.

StealthMark 1.8.8 results for Owl Alpha (LongCat-2.0): benchmark score 25.5 out of 100, General Intelligence 17.2, Visual Intelligence 34.4. Strongest at turning a brief into an original character: 61.6 on Selfie.

Every StealthMark result of Owl Alpha (LongCat-2.0): Selfie 61.6, Beach 54.4, Raccoon 42.6, Newton's Mirror 23.1, Ground Truth 12.9, Mona Lisa 4.1. Not tested yet: Foresight, Latch.

See what it was built for

The case file. First seen 28 Apr 2026, revealed 29 Jun 2026, developer Meituan, context 1,048,576 tokens, input Text, tool calls Yes, native, speed in our tests about 86 tok/s. 28 Apr 2026: Appears with free access, a 1M-token context and no developer’s name. 29 Jun 2026: Meituan names LongCat-2.0 as the model behind Owl Alpha. 12 Jul 2026: The LongCat-2.0 weights follow under the MIT license.

See how it was revealed

Facts as text
Fact Value
First seen 28 Apr 2026
Revealed 29 Jun 2026
Developer Meituan
Context 1,048,576 tokens
Input Text
Tool calls Yes, native
Speed in our tests about 86 tok/s
Access Stealth access ended; available as LongCat-2.0

Run it

The weights are open at meituan-longcat/LongCat-2.0 under MIT. With 1.6 trillion parameters it needs a multi-GPU server, and there's no GGUF yet.


Model guide · StealthMark leaderboard · Follow us
Text and images: CC BY 4.0, stealthmodels.com.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Space using stealth-model/owl-alpha 1