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

Dense decoder transformers

The GPT design: one stack of transformer layers predicting the next token. 308 models on EveryLLM so far.

The dense decoder transformer is the design behind GPT and most chatbots. It is a single stack of identical transformer layers that reads the tokens so far and predicts the next one, with a mask that stops each position from looking ahead. Dense means every parameter is used for every token, unlike in a mixture of experts. GPT-3, PaLM, Llama 3, Chinchilla, Gemma, Phi and OLMo are dense decoders.

Inside each layer

Each layer has two parts. Self-attention lets every token weigh every earlier token and gather what it needs; a feed-forward network then transforms each token's vector on its own. The transformer was introduced by Google researchers in the June 2017 paper Attention Is All You Need, for translation, and OpenAI's GPT-1 (2018) kept only its decoder half. Llama 3.1 405B, released in July 2024, stacks 126 such layers.

Why it won

A decoder trains on any text without labels, because the next token is always the answer to predict, and it scales smoothly: scaling laws published by OpenAI in 2020 and DeepMind in 2022 showed the error falling predictably with more parameters, data and compute. Its weak point is cost. Attention grows with the square of the context length, and every parameter must be read for every token, which is why mixtures of experts became popular.

Famous dense decoder models

Still to be written: OLMo (language model).

Further reading

IBM Granite 🇺🇸 United StatesIBM Granite IBM's decoder-only foundation models, from 3 to 34 billion parametersGeneral · Open weights · 2023 Falcon 3 🇦🇪 United Arab EmiratesFalcon 3 Open foundation models from 1B to 10B parameters by TIIGeneral · Open weights · 2024 StarCoder2 WorldStarCoder2 Open code generation model by BigCodeCode · Open weights · 2024 fnDevstral 🇫🇷 FranceDevstral A 24B agentic coding model built by Mistral AI and All Hands AICode · Open weights · 2025 OpenCoder WorldOpenCoder Open code model family with full training data and pipelineCode · Open weights · 2024 StarCoder WorldStarCoder Open code generation model with 15.5 billion parametersCode · Open weights · 2023 Qwen-72B 🇨🇳 ChinaQwen-72B A 72-billion parameter open-weight language model from Alibaba Cloud.General · Open weights · 2023 Kanana 🇰🇷 South KoreaKanana Bilingual language models for Korean and English by Kakao.General · Open weights · 2025 fnGLM-4.7 🇨🇳 ChinaGLM-4.7 Coding and reasoning model from Z.aiCode · Open weights fnGLM-5.3 🇨🇳 ChinaGLM-5.3 Open-weights coding and cyber-security model from Z.aiCode · Open weights · 2026 MiMo-7B 🇨🇳 ChinaMiMo-7B Open reasoning model from Xiaomi for maths and codeReasoning · Open weights · 2025 imgfnDevstral 2 🇫🇷 FranceDevstral 2 Open-weight coding model family for software agentsCode · Open weights · 2026 Qwen2.5-72B 🇨🇳 ChinaQwen2.5-72B A 72-billion parameter open-weight language model from Alibaba Cloud.General · Open weights · 2024 CodeGen 🇺🇸 United StatesCodeGen Open autoregressive language model for code generationCode · Open weights · 2022 DeepSeek Coder 🇨🇳 ChinaDeepSeek Coder Open-source code generation models from DeepSeekCode · Open weights · 2024 Qwen2 🇨🇳 ChinaQwen2 Open-weight language model series from AlibabaGeneral · Open weights · 2024 Qwen2.5-Coder 🇨🇳 ChinaQwen2.5-Coder Open code-specialised language model series by Alibaba CloudCode · Open weights · 2024 Granite Code 🇺🇸 United StatesGranite Code IBM's open code models from 3B to 34B parameters for coding tasksCode · Open weights · 2024 fnInternLM2 🇨🇳 ChinaInternLM2 Open-source 20 billion parameter language model with long context supportGeneral · Open weights · 2024 fnMiniMax-M2.7 🇨🇳 ChinaMiniMax-M2.7 Open-weight model for coding and agent tasksCode · Open weights Phi-1 🇺🇸 United StatesPhi-1 A small Microsoft model specialised for basic Python codingCode · Open weights · 2023 DeepSeek-V2 🇨🇳 ChinaDeepSeek-V2 Efficient mixture-of-experts language model with 236 billion parametersGeneral · Open weights · 2024

The weekly digest

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