Unlimited Open Source Models

Get Plan
Skip to main content
Available now on ModelsLab · AI Model

Facebook CWMCode Worlds LLM

Try Facebook CWMAPI Documentation

Run CWM Efficiently

Reasoning Built-In

Enable Thinking Tags

Inject <think> tags for step-by-step code reasoning in responses.

32B Parameters

Dense Decoder LLM

Deploy facebook cwm model for advanced code and chat tasks via API.

vLLM Ready

Serve Parallel

Use tensor-parallel-size for fast inference on facebook cwm api.

Examples

See what Facebook CWM can create

Copy any prompt below and try it yourself in the playground.

Recursion Haiku

“Write a haiku about recursion in programming. Use precise code concepts and poetic structure.”

SWE-bench Solve

“Solve this SWE-bench task: Fix bug in Python repo handling async file I/O. Provide diff and explanation.”

LiveCodeBench

“Generate Python solution for LiveCodeBench problem on dynamic programming with memoization. Include tests.”

MATH Proof

“Prove this AIME-level math theorem step-by-step using logical reasoning and LaTeX notation.”

For Developers

A few lines of code.
CWM inference. Two commands.

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

  • Serverless: scales to zero, scales to millions
  • Pay per token, on plans from $21/month
  • Python and JavaScript SDKs, plus REST API
API Documentation 
import requests


response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
  "key": "YOUR_API_KEY",
  "prompt": "",
  "model_id": ""
}
)
print(response.json())

FAQ

Common questions about Facebook CWM

Read the docs

Facebook CWM is a 32-billion-parameter dense decoder-only LLM for code generation. Developed by Meta FAIR. Available on Hugging Face.

Serve with vLLM using facebook/cwm model. Query via OpenAI-compatible /v1/chat/completions endpoint. Enable reasoning with chat_template_kwargs.

Uses CWM License. Request access on Hugging Face for weights. Includes pre-trained, SFT, and post-trained versions.

Facebook CWM is not loaded on ModelsLab at the moment (checked 2026-09-23), so check its model page before you build on it. When it is available it runs on the same OpenAI-compatible chat endpoint as other ModelsLab LLMs, on a ModelsLab plan from $21/month. There is no free tier.

Yes, enables <think> tags by default in vLLM. Disable via extra_body reasoning enabled false. Use specific system prompt.

Hugging Face repos: facebook/cwm, cwm-sft, cwm-pretrain. PyTorch checkpoints from Meta AI resources.

Ready to create?

Start generating with Facebook CWM on ModelsLab.

Try Facebook CWMAPI Documentation