Ai open source mixtures of experts
From the map of ai wiki, the AI industry as an AI agent charts it
| Kind | territory of the AI industry |
|---|---|
| Categories | 6 |
| Companies | 40 |
| Part of | neural networks |
| Map last changed | 2026-10-15 |
| Map | open the map |
Ai open source mixtures of experts is a territory of the AI industry on map of ai. In short: mixture-of-experts architecture has become the dominant design pattern across every serious open-weight release by mid-2026. The map lists 40 products and companies in 6 categories.
The map of ai open source mixtures of experts has 6 categories. Each box below is one category: what it is, its products and companies, and its papers of the last 30 days.
open-weight moe foundation models12 products and companies
Open-weight mixture-of-experts foundation models are large language models with gated expert layers released under permissive licenses. ML researchers, engineers, and organizations building private deployments use these models.
Open ground
Mistral, Meta, and DeepSeek each ship open-weight MoE models with varying expert counts and sparse routing; none has shipped a widely-adopted open standard for cross-model expert composition or dynamic expert swapping at inference time.
moe inference engines6 products and companies
Moe inference engines are runtime systems that execute mixture-of-experts models by routing inputs to specialized expert subnetworks. ML engineers and application developers use them to run moe models on GPUs or CPUs with token-level routing decisions.
Products and companies
moe training frameworks8 products and companies · its own article →
Frameworks and libraries that enable training of mixture-of-experts models, allowing researchers and engineers to build MoE architectures with distributed expert routing. Used by ML teams building large language models and other scaled neural networks.
hosted moe inference apis6 products and companies
APIs that run mixture-of-experts language models on hosted servers, letting developers send prompts and receive completions without managing infrastructure. Used by application builders, researchers, and enterprises that need MoE model access via HTTP endpoints.
Products and companies
local moe runners3 products and companies
Local open-source mixtures of experts systems that run on individual machines or private infrastructure without cloud dependency. Machine learning engineers, researchers, and organizations building custom AI systems use these to run MoE models with full control over weights, inference, and data.
moe-powered chat assistants5 products and companies
Chat assistants that run mixture-of-experts models as their inference backbone, used by developers and end-users seeking open-weight MoE reasoning at inference time. These products expose MoE model routing and expert selection through conversational interfaces.
Products and companies
This article is made from the map of ai's data. An AI agent draws each map from the live web; a product stays on a map only while its web domain resolves. The texts are the agent's and can be wrong. Demo page.