LongCat-2.0

Name: LongCat-2.0
Author: Meituan

Released

2026-06-30

Last refreshed

2026-07-07

Status

Researched 59d ago

Open sourceCommercial use: permittedCodingRAGAgentsLong contextJSON / Tool use

LongCat-2.0 is worth evaluating for coding, rag, and agents when its provider route and context window match the workload.

Use it for

Teams evaluating coding, rag, and agents
Workloads that can use a 1m context window
Buyers comparing 1 tracked provider route

Do not use it for

Vision or document-understanding workloads

Specifications

Family: LongCat
Released: 2026-06-30
Context: 1m
Max output: 128,000
Parameters: 1.6T total; ~48B active
Architecture: Mixture of Experts
Specialization: coding
Openness: Open source
License: MITOSI-approvedCommercial use: permitted
Weights: Available
Code: Unknown
Training: Pretrained

Created by

Meituan

Chinese technology company whose in-house AI team builds the LongCat language-model family.

Beijing, China

Website

Pricing

Output / 1M

$2.95

Input / 1M

$0.750

Cheapest of 1 route · LongCat API Platform · cache read $0.015

Providers(1)

LongCat API Platform

View 1 provider route

Links

Website HuggingFace

About

LongCat-2.0 is Meituan's large-scale MoE language model for coding, long-context, and agentic tasks. The model has 1.6T total parameters, roughly 48B activated per token, native 1M context support, and API access through OpenAI- and Anthropic-compatible LongCat endpoints. Public MIT-licensed safetensors weights are downloadable from Hugging Face, with Transformers, vLLM, and SGLang deployment support.

Top use-case fit: coding, agents, and build tasks

Coding

Q/$ D

1 relevant benchmark in the decision map.

RAG

Included by capability and metadata signals in the decision map.

Agents

Included by capability and metadata signals in the decision map.

Provider price ladder

Compare API pricing across 1 providers for input and output tokens, batch, and cached reads when available.

Provider	Input / 1M	Output / 1M	Cache	Route
LongCat API Platform	$0.750	$2.95	read $0.015	Serverless

Capabilities

ReasoningJSON / Tool usePrompt Caching

Benchmark peer barsfor Coding

SWE-bench ProRank 13 of 46

80.3

79.2

69.2

67.7

59.5

Benchmark scores(4)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.

Benchmark	Score	Version	Evaluation	Source
Terminal-Bench 2.1	70.8	Terminal-Bench 2.1; Code Agent chart; LongCat launch benchmarkObserved 2026-06-30	—	Source
SWE-bench Pro	59.5	SWE-bench Pro; Code Agent chart; LongCat launch benchmarkObserved 2026-06-30	—	Source
SWE-bench Multilingual	77.3	SWE-bench Multilingual; Code Agent chart; LongCat launch benchmarkObserved 2026-06-30	—	Source
BrowseComp	79.9	BrowseComp; Search Agent chart; LongCat launch benchmarkObserved 2026-06-30	—	Source