New Release: Darwin-60B-DUO: Two SOTAs, One Endpoint β€” 88.38% on GPQA Diamond

#2
by SeaWolf-AI - opened

We're excited to release Darwin-60B-DUO, the Darwin family's first DUO model. Take two domain-verified specialists, hide them behind a single OpenAI-compatible endpoint, and let a router decide which one (or both) answers. You see one model, one API β€” but get the best of both.

The number that matters: on the full 198-question GPQA Diamond, Darwin-60B-DUO hits 88.38%. The constituents alone land at 69.70% (Darwin-28B-REASON) and 77.27% (AWAXIS-Think-31B); a naive cascade only reaches 83.84%. The DUO clears them all. Two small specialists, intelligently routed, beat one big generalist on cost and quality. Both are independently verified β€” Darwin-28B-REASON is #3 on the HF GPQA Diamond leaderboard, AWAXIS-Think-31B is #1 on Korea's national K-AI Leaderboard (MSIT).

The brains is a Hybrid-A router picking one of five strategies on the fly. Korean β†’ AWAXIS, English/STEM β†’ Darwin (single-backend, ~70% of traffic at 1Γ— cost). When a Korean answer needs rigorous English reasoning, split_refine fires β€” Darwin drafts, AWAXIS polishes; MCQ/short-answer runs both with self-consistency + cross-verify. Net effective cost: only ~1.3Γ— a single 30B model.

The part the community will care about: the gateway is model-agnostic and Apache-2.0. Point it at any two OpenAI-compatible backends and you've got a DUO in minutes β€” teach router.py when to use which, and parallel calls, response merging, and routing transparency via _duo_route are handled for you. Fork it and tell us what you built.

Painless deploy: docker compose up for both vLLM backends + gateway; FP8 30GB colocates on a single B200/H100. One git clone (120GB). Text-only for now, streaming in v1.1.
Two SOTAs, one endpoint. Come build your own on the Community tab.

πŸ‘‡https://huggingface.co/FINAL-Bench/Darwin-60B-DUO

Looks like it's uploaded in an odd way? Is it the model in the subdirectory awaxis-31b?

Looks like it's uploaded in an odd way? Is it the model in the subdirectory awaxis-31b?

Thanks for checking.

This repo is not a single standard model uploaded in the usual root layout. Darwin-60B-DUO is a DUO/router-style model that orchestrates two underlying 31B-class models. The awaxis-31b subdirectory is one of the routed component models, not the whole model by itself.

So yes, the structure may look unusual if you expect a single model at the repo root. The intended design is closer to a mixture-of-models / routed ensemble, with a custom DarwinDuoOrchestrator wrapper.

I’ll update the README to make this clearer and document which subdirectories are component models and how the DUO router is intended to load them.

Sign up or log in to comment