# Cinq

- **Event:** [GPT-5 Startup Hackathon NYC](https://cerebralvalley.ai/e/gpt5-nyc)
- **When:** Sep 27 at 9:00 AM – Sep 28 at 5:00 PM (EDT)
- **Where:** The Malin Flatiron, New York, NY
- **Team:** [Richard Zhu](https://cerebralvalley.ai/u/rich123)
- **GitHub:** https://github.com/RichardZhu123/autorouter-api
- **Demo video:** https://www.loom.com/share/77be8b1535ca48c5be5669f42b5742aa?sid=b39a1b45-5b8a-4b46-821b-2a90876e0549
- **Gallery:** https://cerebralvalley.ai/e/gpt5-nyc/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/gpt5-nyc/hackathon/gallery/13

Some smaller, specialized models can outperform larger models on specific tasks. What if we have enough coverage from these specialized models to outperform larger models on general tasks? This becomes a routing issue. I present autorouter, a publicly served API, compatible with the OpenAI Python SDK that you can use right now. Achieves state of the art.

---

Markdown version of https://cerebralvalley.ai/e/gpt5-nyc/hackathon/gallery/13. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt