A small team with one strong opinion.
Moamao is an independent product, not a lab and not a platform play. We build one thing: mixture-of-agents orchestration behind a single API call.
The thesis: no single model wins every prompt.
Every frontier model has blind spots, and they don't overlap. The original Mixture-of-Agents study (Wang et al., 2024) showed a panel built entirely from open-source models scoring 65.1% on AlpacaEval 2.0 — ahead of GPT-4 Omni's 57.5%. The gain came from the mixture, not from any single member.
That result is the whole reason Moamao exists. Drafting in parallel, cross-critiquing, and synthesizing measurably beats picking a favorite model and hoping. We turned that loop into one API call with a trace you can audit.
Our principles: what we won't compromise on.
Model-agnostic by design
No provider is privileged. Panels mix models from any vendor, and when a better model ships, it slots into your panel without a migration.
BYOK-first
Your provider keys, your accounts, your rates. We orchestrate; we don't sit between you and your providers' terms.
Every answer auditable
Each run keeps its full trace — every draft, critique, and judge decision. If you can't see how an answer was made, you shouldn't ship it.
Priced in runs, not tokens
You asked one question and got one answer. That's the unit you're billed in, however many model calls happened inside.
We read everything sent to hello@moamao.com. Questions about the beta, the architecture, or whether a mixture fits your workload are all fair game.
Open beta: stop betting everything on one model.
Create a free account, generate an API key, and run your first mixture in minutes — or watch it work in the playground first. BYOK from day one.