LongCat-2.0

Language Meituan Verified Released Jun 30, 2026

LongCat-2.0 is Meituan’s language model for conversations through its official chat service, with options to run it using GPUs or NPUs. Summary written by AI Index from Meituan’s own pages.

  • In Input modalities: Text Out Output modalities: Text
  • Reasoning
  • Reasoning can be off
  • Tool use
  • Open weights

Prices

2 offerings · prices as listed
LongCat API (opens in a new tab) Official Global · USD LongCat-2.0
Conditions/ 1M tokensInputCache readOutput
Standard $0.30$0.006$1.20

Prices from LongCat API: pricing (LongCat-2.0) (opens in a new tab) (checked Nov 9, 2026)

LongCat API (opens in a new tab) Official China · CNY LongCat-2.0
Conditions/ 1M tokensInputCache readOutput
Standard ¥2.00¥0.040¥8.00

Prices from LongCat API: pricing (LongCat-2.0) (opens in a new tab) (checked Nov 9, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Licence
MIT License
Family
longcat
Model ID

Not stated on Meituan’s pages: max input, reasoning efforts, max thinking budget, structured output, built-in tools, knowledge cutoff, aliases.

Run locally

Weights: meituan-longcat/LongCat-2.0
  • Parameters 1.78T
  • Active 48B · 768 experts, 12 per token
  • Native precision BF16
  • Layers 38
  • KV cache per token 43 KB
  • Config positions 256K (config.json max_position_embeddings)
  • API context 977K (maker’s API)
My hardware GeForce RTX 4090 24 GB · 1008 GB/s · 64 GB DDR5 dual channel · 89.6 GB/s
On your hardware · GeForce RTX 4090
Doesn’t fit IQ4_XS (automatic) 8K context
WeightsKV cacheOverheadUsable
Needs
≈ 922.7 GB (weights 878.5 + KV cache 0.3 + overhead 43.9)
Usable
21.6 GB (90% of its memory)
Try
it’s too large for this hardware; see the smallest that runs it below
PrecisionWeightsNeeds at 8KOn your hardwareLongest context
BF16 (native)maker’s file3307.8 GB≈ 3473.5Doesn’t fit—
Q8_01757.0 GB≈ 1845.2Doesn’t fit—
INT8 (GPTQ)maker’s file1918.7 GB≈ 2015.0Doesn’t fit—
FP8maker’s file1910.3 GB≈ 2006.2Doesn’t fit—
Q6_K1358.0 GB≈ 1426.3Doesn’t fit—
Q5_K_M1176.1 GB≈ 1235.3Doesn’t fit—
Q5_01136.9 GB≈ 1194.0Doesn’t fit—
Q4_K_M1004.6 GB≈ 1055.1Doesn’t fit—
Q4_0930.2 GB≈ 977.0Doesn’t fit—
IQ4_XS878.5 GB≈ 922.7Doesn’t fit—
Q3_K_Mquality drops808.2 GB≈ 849.0Doesn’t fit—
IQ3_XXSquality drops632.5 GB≈ 664.5Doesn’t fit—
Q2_Kquality drops620.1 GB≈ 651.4Doesn’t fit—
IQ2_XXSquality drops425.8 GB≈ 447.4Doesn’t fit—

Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.

Compared with other hardware

Each at the highest precision it runs. Click one for details.

Smallest that runs it (Q4_K_M, 8K): no single listed machine; about 15 × 80 GB GPUs.

Run commands

The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.

Memory to fine-tune
Full≈ 26461.4 GBDoesn’t fit your hardware
LoRA≈ 3311.3 GBDoesn’t fit your hardware
QLoRA≈ 934.2 GBDoesn’t fit your hardware

Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.

Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-21. How it’s estimated

Cited scores

From third-party leaderboards, not measured by us

This model isn’t on any leaderboard we cite yet.

  1. Output price cut 60% ¥20.00 → ¥8.00 / 1M tokens · LongCat API, China
  2. Cache read price cut 60% ¥0.10 → ¥0.040 / 1M tokens · LongCat API, China
  3. Input price cut 60% ¥5.00 → ¥2.00 / 1M tokens · LongCat API, China
  4. Output price cut 59.3% $2.95 → $1.20 / 1M tokens · LongCat API
  5. Cache read price cut 60% $0.015 → $0.006 / 1M tokens · LongCat API
  6. Input price cut 60% $0.75 → $0.30 / 1M tokens · LongCat API

Sources

Meituan · China · longcat.chat

Snapshots, the same family and other models from this provider