NOUMENOpen app
NOUMEN

longcat · Open weights

LongCat 2.0 Thinking

Added Jul 21, 2026Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Meituan's LongCat 2.0 Thinking is the reasoning-enabled variant for harder coding, tool use, multi-step reasoning, and long-context agent workflows.

  • Reasoning
  • Tools
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.75
Output
$3.00
Cache read
$0.015

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
131.1K
Average output (7 days)
1.1K tokens
Longer than 73% of 468 models

Side by side

LongCat 2.0 Thinking compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

LongCat 2.0 ThinkinglongcatThis modelQwen3.6 27B TEEQwenRemove
ProviderlongcatQwen
Discovery tierOpen weightsOpen weights
AddedJul 21, 2026Jun 4, 2026
Limits
Context window1MBest262.14K
Max output131.07KBest65.54K
Average output1.1K tokens245 tokens
Pricing, USD per 1M tokens
Input$0.75$0.32Best
Output$3.00$2.70Best
Cache read$0.015—
Performance
Throughput57.2 tok/s—
Time to first token4.0 s—
Request success (24h)100.0%100.0%
Benchmark indexes
Intelligence Index19.119.8Best
Coding Index45.346.6Best
Capabilities
ReasoningYesYes
ToolsYesNo
VisionNoYes
Reasoning effort——

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • LongCat 2.0 Thinking
  • Qwen3.6 27B TEE

Intelligence Index

  • LongCat 2.0 Thinking19.1
  • Qwen3.6 27B TEE19.8

LongCat 2.0 Thinking: Better than 48% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • LongCat 2.0 Thinking: 45.3
    • Qwen3.6 27B TEE: 46.6Top

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • LongCat 2.0 Thinking: 4.5%
    • Qwen3.6 27B TEE: —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • LongCat 2.0 Thinking: 0.5%
    • Qwen3.6 27B TEE: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • LongCat 2.0 Thinking: 739 Elo
    • Qwen3.6 27B TEE: —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • LongCat 2.0 Thinking: 869 Elo
    • Qwen3.6 27B TEE: 961 EloTop

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • LongCat 2.0 Thinking: 6.4%
    • Qwen3.6 27B TEE: —

  • AA-LCRv1.1

    Reasoning across long documents

    • LongCat 2.0 Thinking: 65.0%
    • Qwen3.6 27B TEE: 66.7%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • LongCat 2.0 Thinking: 33.7%Top
    • Qwen3.6 27B TEE: 15.1%

  • GPQA Diamond

    Graduate-level science questions

    • LongCat 2.0 Thinking: 78.0%
    • Qwen3.6 27B TEE: 82.9%Top

  • IFBench

    Following precise output instructions

    • LongCat 2.0 Thinking: —
    • Qwen3.6 27B TEE: 45.7%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • LongCat 2.0 Thinking: 0.0%
    • Qwen3.6 27B TEE: —

  • SciCode

    Scientific computing in Python

    • LongCat 2.0 Thinking: 36.3%
    • Qwen3.6 27B TEE: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • LongCat 2.0 Thinking: —
    • Qwen3.6 27B TEE: 21.2%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • LongCat 2.0 Thinking: —
    • Qwen3.6 27B TEE: 93.6%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • LongCat 2.0 Thinking: 65.0%
    • Qwen3.6 27B TEE: 66.7%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
57.2 tok/s
Faster than 40% of 258 models
Time to first token
4.0 s
Quicker than 44% of 258 models
Prompt cache hits
20.8%
About $0.60 per 1M input tokens with caching
Request success (24h)
100.0%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 100.0%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "longcat-2.0:thinking",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="longcat-2.0:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

longcat

Other longcat models, starting with the same family.