News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM

Mixture of experts

Mixtral, DeepSeek-V3, Kimi K2: many expert networks, a few used for each token. 112 models on EveryLLM so far.

A mixture of experts (MoE) model replaces each dense feed-forward block with many smaller expert networks and a router that sends each token to only a few of them. The model can hold a very large number of parameters while using only a fraction for any one token, so it learns more without costing proportionally more to run. Mixtral, DeepSeek-V3, Qwen3, Llama 4, gpt-oss and Kimi K2 are all built this way.

Total and active parameters

An MoE has two sizes. DeepSeek-V3 (December 2024) holds 671 billion parameters but activates 37 billion for each token. Mixtral 8x7B (December 2023) has eight experts per layer, uses two for each token, and has 46.7 billion parameters in total with 12.9 billion active. Kimi K2 (July 2025) reaches one trillion in total with 32 billion active. All the experts must still sit in memory, so MoE models need large servers even when they are fast.

A long history

The idea dates to 1991, and Google researchers led by Noam Shazeer revived it for language in 2017 with a sparsely gated layer of thousands of experts. Google's Switch Transformer reached 1.6 trillion parameters in 2021. Open MoE models took off with Mixtral, then Databricks' DBRX (March 2024, 132 billion total, 36 billion active), and became the norm for large Chinese open models. Training them is harder, because routers must spread tokens evenly across experts.

Famous mixture of experts models

Further reading

ERNIE 5.5 Flash 🇨🇳 ChinaERNIE 5.5 Flash Efficient sparse MoE language model with 280B total parametersGeneral · Open weights Snowflake Arctic 🇺🇸 United StatesSnowflake Arctic Open enterprise language model with hybrid MoE designGeneral · Open weights · 2024 DeepSeek-Coder-V2 🇨🇳 ChinaDeepSeek-Coder-V2 Open-source mixture-of-experts model for code and mathCode · Open weights · 2024 INTELLECT-3 WorldINTELLECT-3 Open 106B MoE reasoning model from Prime IntellectReasoning · Open weights fnQwen3-Coder 🇨🇳 ChinaQwen3-Coder Agentic coding model with 480 billion parametersCode · Open weights · 2025 imgfnQwen3.6-35B-A3B 🇨🇳 ChinaQwen3.6-35B-A3B Open-weight multimodal coding model with hybrid architectureCode · Open weights · 2026 fnKimi K2 🇨🇳 ChinaKimi K2 Open mixture-of-experts model for agentic tasks and codingGeneral · Open weights · 2025 Qwen3.8-2.4T-A95B 🇨🇳 ChinaQwen3.8-2.4T-A95B Open-weight reasoning model with 2.4 trillion parametersReasoning · Open weights · 2026 imgvidfnKimi K2.6 🇨🇳 ChinaKimi K2.6 Open-source multimodal agentic model by Moonshot AIGeneral · Open weights fnLongCat-2.0 🇨🇳 ChinaLongCat-2.0 A 1.6 trillion parameter MoE language model from Meituan.General · Open weights fnMiniMax-M1 🇨🇳 ChinaMiniMax-M1 Open-weight reasoning model with hybrid attention and long contextReasoning · Open weights · 2025 wavfnSarvam-30B 🇮🇳 IndiaSarvam-30B Open-weight MoE model for Indian languages and codingGeneral · Open weights MiMo-V2.5-Pro 🇨🇳 ChinaMiMo-V2.5-Pro Open-source MoE language model with 1M token contextGeneral · Open weights Qwen2 🇨🇳 ChinaQwen2 Open-weight language model series from AlibabaGeneral · Open weights · 2024 fnMiniMax-M2 🇨🇳 ChinaMiniMax-M2 Open-source MoE model for coding and agentsCode · Open weights DeepSeek-V2 🇨🇳 ChinaDeepSeek-V2 Efficient mixture-of-experts language model with 236 billion parametersGeneral · Open weights · 2024 imgfnLeanstral 1.5 🇫🇷 FranceLeanstral 1.5 An open code agent model built for the Lean 4 proof assistantCode · Open weights

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.