# togethercomputer/MoA

Together Mixture-Of-Agents (MoA) –  65.1% on AlpacaEval with OSS models

Repository: https://github.com/togethercomputer/MoA
Canonical: https://ross.abutalabs.com/products/moa
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-01-07T03:21:19+00:00

## Health v2 (maintenance only)
Score: 24/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 58
- inputs: {"age_days": 820, "days_push": 603, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2970, forks 386 (observed 2026-08-28T04:07:32.722217+00:00)

## What it is
Together AI's Mixture-of-Agents (MoA) is a Python implementation of a layered LLM architecture where multiple open-source models answer prompts and an aggregator model combines their outputs. It achieved 65.1% on AlpacaEval 2.0, surpassing GPT-4 Omni using only open-source models.

## Use cases
- improve response quality by combining multiple open-source LLMs
- build a multi-layer mixture-of-agents pipeline
- run a multi-turn chatbot that aggregates answers from several models
- replicate the MoA paper results on AlpacaEval
- ensemble LLM outputs without using GPT-4

## When to choose
- you want GPT-4-level quality using only open-source models via the Together API
- you need a simple, minimal-code implementation of the MoA technique
- you want to experiment with layered LLM agent architectures

## When to avoid
- you need a production-grade agent framework with tooling and integrations
- you want to avoid multiple LLM API calls and their latency/cost
- you need local model execution rather than the Together API

## Facets
- artifact type: library
- maturity: active
- function: agent-framework, llm-inference, chatbot, cli
- domain: large-language-models, artificial-intelligence
- platform: python, cli
- tags: mixture-of-agents, llm-orchestration, open-source-models, alpacaeval, together-ai, ai-agents

## Member repositories
- togethercomputer/MoA (main) score 24

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:32.722217+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:31:44.561814+00:00, confidence not recorded.
  - readme: https://github.com/togethercomputer/MoA (fetched 2026-08-28T04:07:32.722217+00:00, sha 04cf1d01dca7)
- Data as of 2026-08-30T08:39:29.467469+00:00.