⚡ Quickstart Guide¶

Get up and running with Turing Engine in under 30 seconds.

⚡ Choose Your Setup Path¶

1. Google Colab (Zero Setup)2. Pre-Built Binary Wheel3. Source Build

Run immediately on a free cloud GPU without installing local tools:

Install pre-compiled C++20 AVX2 native wheels directly from GitHub Releases:

# Direct wheel download for macOS Apple Silicon (arm64):
pip install https://github.com/intutic/turing/releases/download/v0.1.8/turing_engine-0.1.8-cp311-cp311-macosx_15_0_arm64.whl

Clone and build the native C++20 extensions locally:

git clone https://github.com/intutic/turing.git
cd turing
pip install -e ".[dev]"
python setup.py build_ext --inplace

💬 1. Instant Terminal Chat¶

Chat directly in your terminal with real pretrained weights:

# Chat with SmolLM2 or DeepSeek-R1:
turing chat --model smollm2
turing chat --model deepseek-r1-1.5b
turing chat --model qwen3-coder-30b

🖥️ 2. Launching the Serving Server¶

Start serving a 70B model or MoE architecture with Subspace acceleration:

turing serve --model deepseek-r1-7b --port 8000 --device auto

Test OpenAI Endpoint¶

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-r1-7b",
    "messages": [{"role": "user", "content": "Explain Subspace Pruning in two sentences."}]
  }'

Test Anthropic Endpoint¶

curl http://localhost:8000/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: turing-local" \
  -d '{
    "model": "claude-3-5-sonnet",
    "max_tokens": 128,
    "messages": [{"role": "user", "content": "Hello Turing Engine!"}]
  }'