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

T-Pro 2.0 🇷🇺 RussiaT-Pro 2.0 Russian-tuned reasoning model based on Qwen 3Reasoning · Open weights Phi-4 🇺🇸 United StatesPhi-4 A 14B Microsoft Research model trained heavily on synthetic data for reasoning.General · Open weights · 2024 Phi-4-mini 🇺🇸 United StatesPhi-4-mini A 3.8B-parameter compact language model from Microsoft for reasoning, maths and codeSmall · Open weights · 2025 Phi-4-reasoning 🇺🇸 United StatesPhi-4-reasoning A 14B open-weight reasoning model fine-tuned from Phi-4Reasoning · Open weights · 2025 Sarvam-M 🇮🇳 IndiaSarvam-M Hybrid reasoning model for Indian languages and EnglishGeneral · Open weights · 2025 fnMagistral 🇫🇷 FranceMagistral Mistral AI's first reasoning model, in open and enterprise versionsReasoning · Open weights · 2025 DeepSeekMath 🇨🇳 ChinaDeepSeekMath Open language model for mathematical reasoningReasoning · Open weights · 2024 fnSmolLM3 🇺🇸 United StatesSmolLM3 Open 3B parameter language model with reasoning modesGeneral · Open weights fnLlemma 🇺🇸 United StatesLlemma Open language model for mathematical reasoningDomain · Open weights · 2023 OLMo 3 🇺🇸 United StatesOLMo 3 Open language models for scientific research by Allen Institute for AIGeneral · Open weights · 2025 fnQwQ-32B 🇨🇳 ChinaQwQ-32B A 32-billion parameter reasoning model from the Qwen series.Reasoning · Open weights · 2025 fnMotif-2 WorldMotif-2 A 12.7 billion parameter reasoning model by Motif Technologies.Reasoning · Open weights EvoLLM-JP 🇯🇵 JapanEvoLLM-JP Experimental Japanese language model created by merging open-source models.General · Open weights · 2024 imgvidfnQwen3-VL 🇨🇳 ChinaQwen3-VL Vision-language model with deep visual reasoning and long video context.Multimodal · Open weights imgvidHyperCLOVA X SEED Think 🇰🇷 South KoreaHyperCLOVA X SEED Think Efficient reasoning model for Korean and multimodal tasksReasoning · Open weights imgEXAONE 4.5 🇰🇷 South KoreaEXAONE 4.5 Open-weight vision-language model from LG AI ResearchMultimodal · Open weights Fathom-R1-14B WorldFathom-R1-14B Low-cost open reasoning model for mathematicsReasoning · Open weights 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 fnSeed-OSS-36B 🇨🇳 ChinaSeed-OSS-36B Open-source 36B parameter language model with controllable reasoning.General · Open weights · 2025 MiMo-7B 🇨🇳 ChinaMiMo-7B Open reasoning model from Xiaomi for maths and codeReasoning · Open weights · 2025 Baichuan-M2 🇨🇳 ChinaBaichuan-M2 Open-source medical reasoning model with 32 billion parameters.Reasoning · Open weights · 2025 K2-Think WorldK2-Think A 32 billion parameter open reasoning modelReasoning · Open weights · 2025 EXAONE Deep 🇰🇷 South KoreaEXAONE Deep Reasoning-focused language models from LG AI ResearchReasoning · Open weights · 2025 imgvidGLM-4.1V-Thinking 🇨🇳 ChinaGLM-4.1V-Thinking Open-source vision-language model with reasoning focusMultimodal · Open weights · 2025 fnGranite 4.2 🇺🇸 United StatesGranite 4.2 IBM's open reasoning language model family, released August 2026Reasoning · Open weights · 2026 DeepSeek-R1 🇨🇳 ChinaDeepSeek-R1 Open-weight reasoning model from ChinaReasoning · 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.