ME
Meituan
Language model · China ·Jul 2026

LongCat 2.0

Language Open Source
1.6T
Parameters
1M
Context
1
Benchmarks

About LongCat 2.0

LongCat 2.0 is an AI model developed by Meituan, in the language category, released in 2026, made available as an open-source model with 1600B params. Meituan's LongCat-2.0 is a 1.6-trillion-parameter Mixture-of-Experts model trained entirely on Chinese ASICs without NVIDIA hardware. Scores 59.5 on SWE-bench Pro and runs at $0.038 per million tokens with free cache hits. Previously served anonymously as "Owl Alpha" on OpenRouter.

On this page you'll find LongCat 2.0's full specifications, including a context window of 1M tokens, 1 benchmark result. Review provider pricing and benchmark scores below, or compare LongCat 2.0 head-to-head with other language models.

→ More models by Meituan → All language models → AI benchmarks leaderboard

Run LongCat 2.0 locally — what it takes

LongCat 2.0 (1,600B) is too large to run on a single consumer or Apple machine. At Q4 its weights alone are about 896 GB — it needs a multi-GPU / datacenter setup:

  • ≈ 4× AMD Instinct MI325X
  • ≈ 6× NVIDIA B200

Lower quantization (Q3/Q2) or CPU + system-RAM offload (e.g. ktransformers) reduces the requirement at the cost of speed.

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Community

Would you run LongCat 2.0 again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

LongCat 2.0 benchmark scores

Compare on the benchmark table →

Scores on standardised evaluations. Higher is better, and the percentile shows where LongCat 2.0 lands among every model we score on that benchmark.

SWE-bench Pro code
top 33% of 24 59.5%

Benchmarks only say so much. Have you run LongCat 2.0 yourself? Tell everyone how it went — what it is good at, where it fell over, and on what hardware.

More models like LongCat 2.0

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is LongCat 2.0? ▶

Meituan's LongCat-2.0 is a 1.6-trillion-parameter Mixture-of-Experts model trained entirely on Chinese ASICs without NVIDIA hardware. Scores 59.5 on SWE-bench Pro and runs at $0.038 per million tokens with free cache hits. Previously served anonymously as "Owl Alpha" on OpenRouter. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created LongCat 2.0? ▶

LongCat 2.0 was developed by Meituan and released in 2026.

Is LongCat 2.0 open source or proprietary? ▶

LongCat 2.0 is released as open source — its weights are publicly available to download, run locally and fine-tune.

What is LongCat 2.0's context window? ▶

LongCat 2.0 supports a context window of 1M tokens, and has 1600B params.

How much does LongCat 2.0 cost? ▶

Pricing for LongCat 2.0 depends on the provider. See the providers table on this page for the latest API rates.

How does LongCat 2.0 perform on benchmarks? ▶

LongCat 2.0 is benchmarked across 1 evaluation on GenAIList, including SWE-bench Pro (59.5%). See the full benchmark table below and compare it with other models.

Meituan LongCat 2.0 is HERE (Real Tests and Review)
NEW 1.6 TRILLION Open Source AI vs Claude & GLM | LongCat-2.0 🫣
LongCat 2.0: N-Grams Beat More Experts
LongCat-2.0 in 90 Seconds — 1.6T Without Nvidia

Know a great video about LongCat 2.0?