MeituanLanguage modelOpen weights

LongCat-2.5

Summary

LongCat-2.5 is Meituan's successor to LongCat-2.0 for tasks that take many steps. It processes text, images and video and responds with text. About 48 billion of its 1.6 trillion parameters activate per token.

The weights are available under the MIT License. The LongCat API has a shared limit of 1,048,576 tokens for input and output, with up to 262,144 output tokens.

Similar models

More models of the same kind as LongCat-2.5, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Active parameters: About 48 billion per token.; Source: Active parameters

Input and output

Input: Text, images and video; Source: Input
Output: Text; Source: Output
API context window: 1,048,576 tokens shared across input and output.; Source: API context window
Maximum API output: 262,144 tokens per request.; Source: Maximum API output

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Unknown; no source
Watermarking technique: Not known; no source

Licence and access

License: MIT; Source: License
Shutdown: None announced; no source

API

API model ID: LongCat-2.5; Source: API model ID

Prices

Limited-time promotional prices: Per 1M tokens: USD 0.30 input, USD 0.006 cached input and USD 1.20 output. The listed CNY rates are 2, 0.04 and 8 for the same categories. Meituan does not give an end date for the promotion.Source

Frequently asked questions about LongCat-2.5

How large is the context window of LongCat-2.5?

1.05 M tokens. That is how much text LongCat-2.5 reads at once, about 1,570 A4 pages.

Page count as a rule of thumb from Artificial Analysis

Context windowTokens the model reads at once · higher is better

LongCat-2.5Rank 12 of 1121.05M

Gemini 4 Carbon1.05M

Gemini 4 Argon1.05M

Muse Wonder1.05M

Kimi K3.1 Pro1.05M

Muse Ember 1.01.05M

Gemini 3.7 Flash1.05M

Can I run LongCat-2.5 locally?

Yes, the weights are open. With 1.6T parameters it needs a lot of memory, more a server than a laptop.

Stated by Meituan

Open weights: yes
License: MIT
Parameters: 1.6T
active per token: 48B

Sources and data date

Type	Evidence and data date
Input	Documented modalities. Retrieved November 13, 2026. Source: LongCat API visual understanding · LongCat API visual understanding (retrieved November 13, 2026) · LongCat API visual understanding · Editorial description reviewed November 13, 2026
Output	Documented modalities. Retrieved November 13, 2026. Source: LongCat API visual understanding · LongCat API visual understanding (retrieved November 13, 2026) · LongCat API visual understanding · Editorial description reviewed November 13, 2026
Additional source	Meituan LongCat-2.5 release (retrieved November 13, 2026)
Additional source	Meituan LongCat-2.5 model card and weights (retrieved November 13, 2026) · Meituan LongCat-2.5 model card and weights · Editorial description reviewed November 13, 2026
Additional source	LongCat API model parameters (retrieved November 13, 2026) · LongCat API model parameters · Editorial description reviewed November 13, 2026
Additional source	LongCat-2.5 API pricing (retrieved November 13, 2026)