Llama3.2 1b trained on data distilled from gpt4o, claude3.5 and claude opus

tools

ollama run catsarethebest/llama3.2-4oClaude

curl http://localhost:11434/api/chat \
  -d '{
    "model": "catsarethebest/llama3.2-4oClaude",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='catsarethebest/llama3.2-4oClaude',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'catsarethebest/llama3.2-4oClaude',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code ollama launch claude --model catsarethebest/llama3.2-4oClaude

Codex ollama launch codex --model catsarethebest/llama3.2-4oClaude

OpenCode ollama launch opencode --model catsarethebest/llama3.2-4oClaude

OpenClaw ollama launch openclaw --model catsarethebest/llama3.2-4oClaude

Models

View all →

Name

2 models

Size

Context

Input

llama3.2-4oClaude:latest

2.5GB · 128K context window · Text · 10 months ago

llama3.2-4oClaude:latest

2.5GB

128K

Text

llama3.2-4oClaude:o4-mini

2.5GB · 128K context window · Text · 7 months ago

llama3.2-4oClaude:o4-mini

2.5GB

128K

Text

Readme

Llama 3.2 4o Claude

The catsarethebest/llama3.2-4oClaude:latest model was trained on a mix of datasets distilled from GPT-4o, Claude 3.5, and Claude 3.5 Opus.

The catsarethebest/llama3.2-4oClaude:o4-mini version is a fine-tuned variant of :latest, trained further on a smaller, fully synthetic dataset distilled from o4-mini.