LongCat-2.0 Now on SiliconFlow: 1.6T MoE, Native 1M Context, Built for Agentic Coding

TL;DR：

• LongCat-2.0 is now live on SiliconFlow Meituan LongCat's next-generation trillion-parameter flagship model, and it's the full model behind "Owl Alpha", the anonymous model that reached Top 3 globally by daily volume on OpenRouter.

• Features: 1.6T total parameters with ~48B activated per token, native 1M-token context, purpose-built for agentic coding.

• Cost on SiliconFlow:

  • $0.75 / M input · $2.95 / M output · $0.015 / M cache read

• Get started: SiliconFlow provides OpenAI- and Anthropic-compatible APIs for LongCat-2.0, so it drops straight into Claude Code, Cline, Hermes Agent, OpenCode, and your existing tools in minutes

  
  

Some of you have already been trusting a model without knowing its name. "Owl Alpha," the anonymous model on OpenRouter, was actually a preview of LongCat-2.0 — and it reached Top 3 globally by daily volume, ranking #1 on Hermes Agent, #2 on Claude Code, and #3 on OpenClaw by monthly volume.

Meituan LongCat has now shipped the full model behind it, with architecture-level upgrades across the board. LongCat-2.0 is now live on SiliconFlow.

Why LongCat-2.0 stands out

• LongCat Sparse Attention (LSA): scales efficiently for 1M-context tokens. Compute for long-context processing drops from quadratic to linear, so the model can still precisely locate the right information even at the full 1M-token window.

• Zero-Compute Experts + ScMoE: dynamic activation of 33B–56B parameters per token, zero wasted compute. Simple tokens skip the extra compute, while complex ones automatically draw on more — so you're not paying for capacity you don't need.

• MOPD: three specialized expert groups (Agent / Reasoning / Interaction), gate-routed per task. Each group is trained independently and dynamically scheduled at inference rather than simply merged, giving the model strong, balanced performance across coding, reasoning, and interaction.

For the full technical details behind LongCat-2.0's architecture and training, see Meituan LongCat's official blog.

Benchmark performance

LongCat-2.0 leads on SWE-bench Pro and holds its own across FORTE, RWSearch, and BrowseComp — agentic coding and general-agent workloads that go well beyond a single benchmark.

Run LongCat-2.0 with Your Existing Tools

Benchmarks and backstory can only tell you so much — the fastest way to know if LongCat-2.0 fits your workflow is to run it yourself. And that only takes minutes.

Try it first, no setup

Open the SiliconFlow Playground and chat with LongCat-2.0 right in your browser.

Tune temperature and top-p, or run it side by side against GLM-5.2 or DeepSeek-V4-Pro on the same coding prompt to see how it performs.

Plug into your tools

SiliconFlow APIs are both OpenAI- and Anthropic-compatible, so LongCat-2.0 drops straight into any tool that supports a custom provider — same SDK, same code, just a new base URL and model string. Live in minutes.

What you need to connect:

Get Started Immediately

1. Build: Try LongCat-2.0 on SiliconFlow through the playground.

2. Integrate: Use our OpenAI-compatible API. Explore the full API specifications in the SiliconFlow API documentation.

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "meituan-longcat/LongCat-2.0",
    "stream": True,
    "max_tokens": 4096,
    "temperature": 0.7,
    "top_p": 0.95,
    "messages": [
        {
            "role": "user",
            "content": "Review this module and flag any API compatibility risks."
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.request("POST", url, json=payload, headers=headers)

print(response.text)

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "meituan-longcat/LongCat-2.0",
    "stream": True,
    "max_tokens": 4096,
    "temperature": 0.7,
    "top_p": 0.95,
    "messages": [
        {
            "role": "user",
            "content": "Review this module and flag any API compatibility risks."
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.request("POST", url, json=payload, headers=headers)

print(response.text)

import requests

url = "https://api.siliconflow.com/v1/chat/completions"

payload = {
    "model": "meituan-longcat/LongCat-2.0",
    "stream": True,
    "max_tokens": 4096,
    "temperature": 0.7,
    "top_p": 0.95,
    "messages": [
        {
            "role": "user",
            "content": "Review this module and flag any API compatibility risks."
        }
    ]
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.request("POST", url, json=payload, headers=headers)

print(response.text)

Business or Sales Inquiries →

Join our Discord community now →

Follow us on X for the latest updates →

Explore all available models on SiliconFlow →