Using Token Factory Models with Coding Agents
Token Factory has integrations with many coding agents and allows you to run state of the art open source models.
Checkout all coding integrations for instrcutions
Cursor
demo:
Cline
demo:
Claude Code
Using Claude Code with Proxy Server
Follow the instructions there.
Recommended coding models on Token Factory (as of March 2026)
- zai-org/GLM-5.1
- zai-org/GLM-4.7-FP8
- Qwen/Qwen3-Coder-480B-A35B-Instruct
Sample .env file for Qwen/Qwen3-Coder-480B-A35B-Instruct
OPENAI_API_KEY="your key goes here"
ANTHROPIC_API_KEY="dummy"
IGNORE_CLIENT_API_KEY=true
OPENAI_BASE_URL=https://api.tokenfactory.nebius.com/v1
BIG_MODEL=Qwen/Qwen3-Coder-480B-A35B-Instruct
MIDDLE_MODEL=Qwen/Qwen3-Coder-480B-A35B-Instruct
SMALL_MODEL=Qwen/Qwen3-Coder-480B-A35B-Instruct
VISION_MODEL=Qwen/Qwen2.5-VL-72B-Instruct
# Optional: explicit context limits (tokens) for safer max_tokens auto-capping
BIG_MODEL_CONTEXT_LIMIT=204800
MIDDLE_MODEL_CONTEXT_LIMIT=204800
SMALL_MODEL_CONTEXT_LIMIT=204800
VISION_MODEL_CONTEXT_LIMIT=32768
STRIP_IMAGE_CONTEXT=true
HOST=0.0.0.0
PORT=8083
LOG_LEVEL=INFO #DEBUG
## testing this
MAX_TOKENS_LIMIT=64000
#MIN_TOKENS_LIMIT=4096
REQUEST_TIMEOUT=90
MAX_RETRIES=2