Summer 1956In the summer of 1956, a small group of scientists gathered at Dartmouth College for the Dartmouth Summer Research Project on Artificial Intelligence.
1956Carnegie Mellon University Archives preserves the September 1956 article by Allen Newell and Herbert A. Simon titled 'The Logic Theory Machine.
1958Frank Rosenblatt published The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain in Psychological Review.
July 31, 1959Arthur L. Samuel published 'Some Studies in Machine Learning Using the Game of Checkers' in the IBM Journal of Research and Development on July 31, 1959.
April 1960John McCarthy published 'Recursive Functions of Symbolic Expressions and Their Computation by Machine, Part I' in April 1960.
3 May 1960Herbert Gelernter, J. R. Hansen and D. W. Loveland published a conference paper documenting empirical work with an IBM 704 geometry theorem-proving machine.
15 December 1964Joshua Lederberg's DENDRAL-64 technical report, Part I, was published on December 15, 1964.
1965MIT Libraries preserves a 1965 archival item described as a complete printout of the source code listing of ELIZA, written in the MAD-SLIP programming language.
April 5, 1965On April 5, 1965, Edward A. Feigenbaum and R. W. Watson issued An Initial Problem Statement for a Machine Induction Research Project at the Stanford Artificial Intelligence Project.
1966SRI began the Shakey research program in 1966.
January 1966Joseph Weizenbaum published ELIZA—A Computer Program for the Study of Natural Language Communication Between Man and Machine in Communications of the ACM in January 1966.
1967J. MacQueen published 'Some methods for classification and analysis of multivariate observations' in the Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics.
1970Terry Winograd later documented that he developed SHRDLU at MIT between 1968 and 1970.
1972In a later first-person account, Terry Winograd states that his original SHRDLU dissertation was Winograd (1971) and that Winograd (1972) was a rewritten version.
1973Waseda University records that Professor Ichiro Kato's team completed WABOT-1 in 1973.
1974Edward H. Shortliffe's Stanford thesis documented MYCIN as a rule-based computer program designed to advise physicians regarding antimicrobial therapy selection.
August 1975In August 1975, Edward H. Shortliffe, Randall Davis, Stanton G. Axline, Bruce G. Buchanan, C. Cordell Green, and Stanley N. Cohen published a paper describing progress on MYCIN.
1977A. P. Dempster, N. M. Laird, and D. B. Rubin published "Maximum Likelihood from Incomplete Data Via the EM Algorithm" in the Journal of the Royal Statistical Society: Series B in 1977.
April 1980Kunihiko Fukushima published Neocognitron.
January 1982Teuvo Kohonen published 'Self-organized formation of topologically correct feature maps' in Biological Cybernetics in January 1982.
April 1982John J. Hopfield published Neural networks and physical systems with emergent collective computational abilities in the Proceedings of the National Academy of Sciences in April 1982.
January 1985David H. Ackley, Geoffrey E. Hinton, and Terrence J. Sejnowski published "A Learning Algorithm for Boltzmann Machines" in Cognitive Science in January 1985.
1986J. R. Quinlan published 'Induction of decision trees' in Machine Learning in 1986.
9 October 1986David E. Rumelhart, Geoffrey E. Hinton and Ronald J. Williams published Learning representations by back-propagating errors in Nature on 9 October 1986.
1988Richard S. Sutton published 'Learning to Predict by the Methods of Temporal Differences' in Machine Learning in 1988.
May 1989Christopher John Cornish Hellaby Watkins submitted his University of Cambridge PhD thesis 'Learning from Delayed Rewards' in May 1989.
May 1992Christopher J. C. H. Watkins and Peter Dayan published Q-learning in Machine Learning in May 1992.
22 October 1993Gerald Tesauro published a TD-Gammon paper for AAAI-FS 1993 on 22 October 1993.
1995Yoav Freund and Robert E. Schapire published "A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting" in the 1995 EuroCOLT proceedings.
September 1995Corinna Cortes and Vladimir Vapnik published Support-Vector Networks in the September 1995 issue of Machine Learning.
11 May 1997IBM's Deep Blue defeated reigning world chess champion Garry Kasparov in a six-game match completed on 11 May 1997, winning 3.5 to 2.5 under standard tournament time controls.
15 November 1997Sepp Hochreiter and Jürgen Schmidhuber published Long Short-Term Memory in Neural Computation on 15 November 1997.
30 November 1998Yann LeCun, Leon Bottou, Yoshua Bengio, and Patrick Haffner published Gradient-Based Learning Applied to Document Recognition in Proceedings of the IEEE.
September 1, 1999Leo Breiman's University of California, Berkeley Statistics technical report 'Random Forests--Random Features' is dated September 1, 1999.
October 2001Leo Breiman published "Random Forests" in Machine Learning in October 2001.
February 2003Yoshua Bengio, Réjean Ducharme, Pascal Vincent, and Christian Jauvin published A Neural Probabilistic Language Model in the Journal of Machine Learning Research in February 2003.
January 2008The authors of the SciPy 2010 paper 'Theano: A CPU and GPU Math Compiler in Python' state that Theano development and use began in January 2008.
2009Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei published ImageNet.
1 February 2010The scikit-learn project history records 1 February 2010 as the date of its first public release, after leadership of the project passed in 2010 to Fabian Pedregosa, Gaël Varoquaux.
2011Ronan Collobert's publication record and Idiap's institutional publication record identify the paper 'Torch7: A Matlab-like Environment for Machine Learning' by Ronan Collobert.
16 February 2011IBM's Watson question-answering system defeated Jeopardy! champions Ken Jennings and Brad Rutter in a televised contest conducted over February 14–16, 2011.
2012Alex Krizhevsky, Ilya Sutskever and Geoffrey E. Hinton published ImageNet Classification with Deep Convolutional Neural Networks in the 2012 Neural Information Processing Systems.
July 3, 2012Geoffrey E. Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever.
January 16, 2013Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean submitted Efficient Estimation of Word Representations in Vector Space to arXiv on January 16, 2013.
19 December 2013Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra.
20 December 2013Diederik P. Kingma and Max Welling submitted 'Auto-Encoding Variational Bayes' on December 20, 2013.
2014Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville.
March 2014The official XGBoost repository preserves a v0.1 tag whose referenced commit is dated March 2014.
20 June 2014The Caffe authors recorded the framework in the paper 'Caffe: Convolutional Architecture for Fast Feature Embedding' on June 20, 2014.
August 2014The Stanford NLP Group records August 2014 as the original release of GloVe, prepared by Jeffrey Pennington.
1 September 2014Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio submitted 'Neural Machine Translation by Jointly Learning to Align and Translate' on September 1, 2014.
September 4, 2014Karen Simonyan and Andrew Zisserman submitted Very Deep Convolutional Networks for Large-Scale Image Recognition to arXiv on September 4, 2014.
10 September 2014Ilya Sutskever, Oriol Vinyals, and Quoc V. Le submitted 'Sequence to Sequence Learning with Neural Networks' on September 10, 2014.
17 September 2014Christian Szegedy and coauthors submitted 'Going Deeper with Convolutions' on September 17, 2014.
December 22, 2014Diederik P. Kingma and Jimmy Ba submitted 'Adam: A Method for Stochastic Optimization' on December 22, 2014.
February 11, 2015Sergey Ioffe and Christian Szegedy submitted Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift on February 11, 2015.
February 25, 2015Volodymyr Mnih and colleagues published Human-level control through deep reinforcement learning online in Nature on February 25, 2015.
March 2015François Chollet's first-party Keras 1.0 announcement states that Keras was initially released in late March 2015.
March 12, 2015Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli submitted 'Deep Unsupervised Learning using Nonequilibrium Thermodynamics' on March 12, 2015.
18 May 2015Olaf Ronneberger, Philipp Fischer, and Thomas Brox submitted 'U-Net: Convolutional Networks for Biomedical Image Segmentation' on May 18, 2015.
June 8, 2015Joseph Redmon, Santosh Divvala, Ross Girshick and Ali Farhadi submitted You Only Look Once: Unified, Real-Time Object Detection to arXiv on June 8, 2015.
9 June 2015Preferred Infrastructure and Preferred Networks announced on June 9, 2015 that they had developed the Chainer deep-learning framework and released it as open-source software.
9 November 2015Google announced TensorFlow on November 9, 2015 and stated that it was open-sourcing the machine-learning system that day.
3 December 2015The MXNet authors submitted the technical paper 'MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems' on December 3, 2015.
December 10, 2015Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun submitted Deep Residual Learning for Image Recognition on December 10, 2015.
2016PaddlePaddle's official project documentation and Baidu's corporate reporting state that Baidu open-sourced the PaddlePaddle deep-learning framework in 2016.
25 January 2016Microsoft's CNTK project history records that the CNTK repository on GitHub went public on January 25, 2016, together with a new CNTK website.
January 27, 2016David Silver and colleagues published Mastering the game of Go with deep neural networks and tree search online in Nature on January 27, 2016.
February 4, 2016Volodymyr Mnih and collaborators submitted 'Asynchronous Methods for Deep Reinforcement Learning' on February 4, 2016.
March 9, 2016Tianqi Chen and Carlos Guestrin submitted XGBoost: A Scalable Tree Boosting System to arXiv on March 9, 2016.
July 21, 2016Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E. Hinton submitted Layer Normalization on July 21, 2016.
8 September 2016Google DeepMind publicly presented WaveNet on September 8, 2016 as a deep generative model of raw audio waveforms.
19 January 2017The PyTorch team retrospectively marked January 19, 2017 as the date PyTorch was released publicly.
April 17, 2017Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto and Hartwig Adam submitted MobileNets.
2017Vaswani and colleagues introduced the Transformer architecture in Attention Is All You Need.
6 September 2017Amazon Web Services announced on September 6, 2017 the availability of Apache MXNet 0.11 and described it as MXNet's first official release as an Apache incubating project.
7 September 2017Microsoft announced the Open Neural Network Exchange (ONNX) format on 7 September 2017 in conjunction with Facebook.
December 2017Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu published "LightGBM: A Highly Efficient Gradient Boosting Decision Tree" in December 2017.
December 5, 2017David Silver and colleagues submitted Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm to arXiv on December 5, 2017.
2018JAX's official repository citation metadata states that 2018 corresponds to the project's open-source release.
15 February 2018Matthew E. Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer submitted 'Deep contextualized word representations' on February 15, 2018.
June 11, 2018OpenAI published Improving language understanding with unsupervised learning on June 11, 2018, describing and releasing a scalable.
October 11, 2018Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova submitted BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding on October 11, 2018.
2 November 2018Google AI announced the open-source release of BERT on November 2, 2018. The release included TensorFlow source code and pre-trained language representation models.
January 9, 2019Zihang Dai, Zhilin Yang, Yiming Yang, Jaime Carbonell, Quoc V. Le, and Ruslan Salakhutdinov submitted 'Transformer-XL: Attentive Language Models beyond a Fixed-Length Context' on January 9.
February 14, 2019OpenAI published Better Language Models and Their Implications on February 14, 2019.
April 19, 2019Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, and Hua Wu submitted the paper 'ERNIE.
23 October 2019Google researchers submitted the T5 paper, 'Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer,' on October 23, 2019.
November 19, 2019Julian Schrittwieser and colleagues submitted the paper Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model to arXiv on November 19, 2019.
March 13, 2020Iroro Orife and collaborators submitted 'Masakhane -- Machine Translation For Africa' on March 13, 2020.
28 March 2020Huawei announced on March 28, 2020 that its all-scenario AI computing framework MindSpore had been officially open-sourced.
May 22, 2020Patrick Lewis and coauthors submitted 'Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks' on May 22, 2020.
May 28, 2020Tom B. Brown and coauthors submitted Language Models are Few-Shot Learners on May 28, 2020.
June 19, 2020Jonathan Ho, Ajay Jain, and Pieter Abbeel submitted 'Denoising Diffusion Probabilistic Models' to arXiv on June 19, 2020.
October 22, 2020Alexey Dosovitskiy and coauthors submitted 'An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale' on October 22, 2020.
January 5, 2021OpenAI published its CLIP milestone on January 5, 2021.
26 April 2021The PanGu-α authors submitted the paper "PanGu-α: Large-scale Autoregressive Pretrained Chinese Language Models with Auto-parallel Computation" on April 26, 2021.
18 May 2021Google introduced LaMDA, short for Language Model for Dialogue Applications, in an official Google I/O research announcement on May 18, 2021.
June 17, 2021Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen submitted 'LoRA.
July 15, 2021John Jumper and colleagues published Highly accurate protein structure prediction with AlphaFold in Nature on July 15, 2021.
29 March 2022DeepMind researchers introduced Chinchilla, a 70-billion-parameter language model, in the paper Training Compute-Optimal Large Language Models submitted to arXiv on March 29, 2022.
5 April 2022Google researchers introduced the Pathways Language Model (PaLM), a 540-billion-parameter densely activated Transformer language model, in a paper submitted to arXiv on April 5, 2022.
28 April 2022DeepMind introduced Flamingo, a visual language model for few-shot learning across multimodal tasks, in an official research post on April 28, 2022.
3 May 2022Meta AI announced on May 3, 2022 that it was sharing Open Pretrained Transformer (OPT-175B), a 175-billion-parameter language model.
12 July 2022The BigScience collaboration publicly released BLOOM on July 12, 2022.
22 August 2022Stability AI publicly released Stable Diffusion on August 22, 2022 following an earlier researcher release, making model weights, a model card and code available for public use.
30 November 2022OpenAI publicly introduced ChatGPT on November 30, 2022 as a conversational model released for research preview and user feedback.
24 February 2023Meta AI announced LLaMA on February 24, 2023 as a collection of foundation language models offered in 7B, 13B, 33B, and 65B parameter sizes.
March 14, 2023OpenAI publicly introduced GPT-4 on March 14, 2023. OpenAI described GPT-4 as a large multimodal model capable of accepting image and text inputs and producing text outputs.
14 March 2023Anthropic announced Claude more broadly on March 14, 2023 after a closed alpha with partners, describing it as a next-generation AI assistant available through a chat interface and API.
2023Abu Dhabi's Technology Innovation Institute announced the launch of Falcon LLM on March 15, 2023.
16 March 2023Baidu introduced ERNIE Bot on March 16, 2023 as a new-generation large language model and generative AI product.
April 16, 2023Ramon Pires, Hugo Abonizio, Thales Sales Almeida and Rodrigo Nogueira submitted Sabiá: Portuguese Large Language Models to arXiv on April 16, 2023.
2023Meta announced the availability of Llama 2 on July 18, 2023 as the next generation of its large language model family.
2023NAVER introduced HyperCLOVA X at its DAN 23 conference on August 24, 2023, alongside CLOVA X and the generative AI search service CUE: built on the model.
30 August 2023Inception, Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), and Cerebras Systems announced the open release of Jais on August 30, 2023.
2023Mistral AI released Mistral 7B on September 27, 2023, describing it as a 7.3-billion-parameter language model released under the Apache 2.0 license.
November 2023AI Singapore states that SEA-LION (Southeast Asian Languages in One Network), its family of open large language models developed to better represent Southeast Asia's diverse contexts.
2023Google and Google DeepMind introduced Gemini on December 6, 2023, with Gemini 1.0 optimized in Ultra, Pro and Nano sizes and designed as a multimodal model family.
11 December 2023Mistral AI released Mixtral 8x7B on December 11, 2023, describing it as a sparse mixture-of-experts model with open weights and licensing it under Apache 2.0.
2024SCB 10X announced Typhoon on January 29, 2024 as a large language model optimized for Thai.
13 February 2024Cohere For AI announced the public launch of Aya on February 13, 2024.
15 February 2024Google announced Gemini 1.5 on February 15, 2024 and began a limited early preview of Gemini 1.5 Pro for developers and enterprise customers through AI Studio and Vertex AI.
February 21, 2024Google introduced Gemma on February 21, 2024 as a family of lightweight open models, releasing Gemma 2B and Gemma 7B in pretrained and instruction-tuned variants.
4 March 2024Anthropic announced the Claude 3 model family on March 4, 2024 and made Claude 3 Opus and Claude 3 Sonnet available that day through claude.ai and the Claude API.
11 March 2024Cohere introduced Command R on March 11, 2024 as a generative model aimed at scalable production workloads, with emphasis on retrieval-augmented generation, tool use, and long-context tasks.
13 March 2024Anthropic released Claude 3 Haiku on March 13, 2024 and stated that it was available through the Claude API and to Claude Pro subscribers on claude.ai.
4 April 2024Cohere introduced Command R+ on April 4, 2024 as a RAG-optimized model for enterprise-grade workloads and stated that it was available first on Microsoft Azure.
17 April 2024Mistral AI released Mixtral 8x22B on April 17, 2024, describing it as a sparse mixture-of-experts model with 141 billion total parameters, 39 billion active parameters.
18 April 2024Meta released the initial Llama 3 models on April 18, 2024. The release included pretrained and instruction-fine-tuned 8B and 70B language models made available for broad use.
14 May 2024Google announced on May 14, 2024 that Gemini 1.5 Pro was being brought to Gemini Advanced subscribers, with availability stated for more than 150 countries and over 35 languages.
14 May 2024Google announced on May 14, 2024 that Gemini 1.5 Pro was available in public preview in Google AI Studio and Vertex AI, alongside Gemini 1.5 Flash.
23 May 2024Cohere For AI announced Aya 23 on May 23, 2024 as a family of multilingual generative language models covering 23 languages, with 8B and 35B parameter open-weight models.
24 May 2024Google Cloud's Vertex AI release notes record Gemini 1.5 Pro (gemini-1.5-pro-001) as generally available on May 24, 2024, alongside Gemini 1.5 Flash (gemini-1.5-flash-001).
29 May 2024Mistral AI introduced Codestral on May 29, 2024 as its first code model, an open-weight 22-billion-parameter model designed for code generation and fill-in-the-middle tasks.
2024The Qwen Team announced Qwen2 on June 7, 2024, releasing pretrained and instruction-tuned models in five sizes and making the models available through Hugging Face and ModelScope.
11 June 2024Google released the RecurrentGemma 9B variant on June 11, 2024.
21 June 2024Anthropic launched Claude 3.5 Sonnet on June 21, 2024 and stated that it was available on Claude.ai, the Claude iOS app, the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI.
27 June 2024Google announced the initial release of Gemma 2 on June 27, 2024, making 9B and 27B parameter models available to researchers and developers.
18 July 2024Mistral AI announced the release of Mistral NeMo on July 18, 2024, a 12B model built in collaboration with NVIDIA.
23 July 2024Meta announced and publicly released the Llama 3.1 model collection on July 23, 2024.
24 July 2024Mistral AI released Mistral Large 2 on July 24, 2024, a 123B-parameter model with a 128k context window and open weights under the Mistral Research License.
31 July 2024Google released the Gemma 2 2B model on July 31, 2024, making its model weights available through Kaggle, Hugging Face.
31 July 2024Google DeepMind announced Gemma Scope on July 31, 2024 as an open suite of sparse autoencoders for language-model interpretability.
31 July 2024Google's official Gemma release ledger records July 31, 2024 as the initial release of ShieldGemma.
2024Sarvam AI introduced Sarvam 2B on August 13, 2024 as an open-source 2-billion-parameter small Indic language model within the launch of its full-stack generative AI platform for India.
30 August 2024Cohere announced refreshed versions of Command R and Command R+ on August 30, 2024, exposed through the API as command-r-08-2024 and command-r-plus-08-2024.
2024The InkubaLM paper was submitted to arXiv on August 30, 2024 and explicitly introduced InkubaLM, predating the later October 25, 2024 Lelapa AI blog post.
12 September 2024Google released DataGemma in 2B size on September 12, 2024.
September 19, 2024The Qwen Team publicly introduced Qwen2.5 on September 19, 2024 and described the release as including Qwen2.5 language models alongside Qwen2.5-Coder and Qwen2.5-Math model families.
25 September 2024Meta announced Llama 3.2 on September 25, 2024, including 11B and 90B vision models and lightweight 1B and 3B text-only models.
15 October 2024Google released Gemma-APS in 2B and 7B sizes on October 15, 2024.
22 October 2024Anthropic announced an upgraded Claude 3.5 Sonnet on October 22, 2024 and introduced computer use in public beta.
24 October 2024Cohere For AI announced Aya Expanse on October 24, 2024 as a multilingual family of 8B and 32B parameter models supporting 23 languages.
2024Preferred Networks announced the launch of PLaMo Prime on December 2, 2024 as the flagship model in its PLaMo large-language-model line.
5 December 2024Google's Gemma release documentation records the release of PaliGemma 2 on December 5, 2024, in 3B, 10B, and 28B parameter sizes.
6 December 2024Meta released Llama 3.3 70B Instruct on December 6, 2024. Meta's official model card records that release date and identifies Llama 3.3 as a multilingual, text-only model.
11 December 2024Google announced Gemini 2.0 on December 11, 2024 and released an experimental version of Gemini 2.0 Flash, the first model in the Gemini 2.0 family.
13 December 2024Cohere announced Command R7B on December 13, 2024 as the smallest and fastest model in its R family of enterprise-focused large language models, with a 128K context window.
26 December 2024DeepSeek introduced DeepSeek-V3 on December 26, 2024.
January 20, 2025DeepSeek publicly released DeepSeek-R1 on January 20, 2025.
30 January 2025Mistral AI introduced Mistral Small 3 on January 30, 2025 as a latency-optimized 24B-parameter model released under the Apache 2.0 license.
19 February 2025xAI introduced Grok 3 and Grok 3 mini in beta on February 19, 2025.
19 February 2025Google's Gemma release documentation records the release of PaliGemma 2 mix on February 19, 2025, in 3B, 10B, and 28B parameter sizes.
24 February 2025Anthropic announced Claude 3.7 Sonnet on February 24, 2025 and made it available across Claude plans and developer platforms.
27 February 2025OpenAI released GPT-4.5 as a research preview on February 27, 2025.
4 March 2025Cohere Labs announced Aya Vision on March 4, 2025 as a multilingual multimodal model supporting text-and-image tasks across 23 languages.
10 March 2025Google's Gemma release documentation records the release of Gemma 3 on March 10, 2025, in 1B, 4B, 12B, and 27B parameter sizes.
10 March 2025Google's official Gemma release ledger records the release of ShieldGemma 2 on March 10, 2025.
13 March 2025Cohere released Command A on March 13, 2025 as an enterprise-focused generative model.
13 March 2025Ai2 released OLMo 2 32B on March 13, 2025 as the largest model in the OLMo 2 family.
17 March 2025Mistral AI announced Mistral Small 3.1 on March 17, 2025 and made the model available under the Apache 2.0 license.
25 March 2025Google introduced Gemini 2.5 on March 25, 2025, beginning with Gemini 2.5 Pro Experimental.
25 March 2025Google introduced TxGemma on March 25, 2025 as a collection of open models designed to support therapeutic-development tasks.
5 April 2025Meta announced Llama 4 Scout and Llama 4 Maverick on April 5, 2025 and made both models available for download that day.
14 April 2025Google announced progress on DolphinGemma on April 14, 2025.
14 April 2025OpenAI launched GPT-4.1, GPT-4.1 mini and GPT-4.1 nano in the API on April 14, 2025.
16 April 2025OpenAI released o3 and o4-mini on April 16, 2025.
29 April 2025The Qwen team announced the release of Qwen3 on April 29, 2025.
7 May 2025Mistral AI announced Mistral Medium 3 on May 7, 2025.
20 May 2025Google announced the preview of Gemma 3n on May 20, 2025 as a mobile-first open model architecture intended for efficient multimodal AI on phones, tablets, and laptops.
21 May 2025Mistral AI introduced and released Devstral on May 21, 2025 for agentic software-engineering tasks.
22 May 2025Anthropic introduced Claude Opus 4 and Claude Sonnet 4 on May 22, 2025.
10 June 2025Mistral AI announced Magistral on June 10, 2025 as its first reasoning-model family.
17 June 2025Google released stable generally available versions of Gemini 2.5 Pro and Gemini 2.5 Flash on June 17, 2025.
26 June 2025Google's Gemma release ledger records June 26, 2025 as the release of Gemma 3n in E2B and E4B sizes.
9 July 2025xAI released Grok 4 on July 9, 2025 and introduced Grok 4 Heavy.
5 August 2025Anthropic released Claude Opus 4.1 on August 5, 2025.
7 August 2025OpenAI introduced GPT-5 on August 7, 2025.
14 August 2025Google introduced Gemma 3 270M on August 14, 2025 as a compact 270-million-parameter model designed for task-specific fine-tuning and efficient on-device use.
4 September 2025Google introduced EmbeddingGemma on September 4, 2025 as an open 308-million-parameter embedding model designed for efficient on-device text embeddings.
13 September 2025Google's Gemma release ledger records September 13, 2025 as the release of VaultGemma in 1B parameter size.
19 September 2025xAI introduced Grok 4 Fast on September 19, 2025.
29 September 2025Anthropic released Claude Sonnet 4.5 on September 29, 2025.
15 October 2025Anthropic released Claude Haiku 4.5 on October 15, 2025.
13 November 2025OpenAI released GPT-5.1 on its API platform on November 13, 2025.
17 November 2025xAI introduced Grok 4.1 on November 17, 2025 and made it available across Grok surfaces.
18 November 2025Google announced the release of Gemini 3 on November 18, 2025, beginning the Gemini 3 era with Gemini 3 Pro in preview and availability across Google products and developer platforms.
20 November 2025Ai2 released Olmo 3 on November 20, 2025 together with a documented model flow, open models, data, and training artifacts.
24 November 2025Anthropic released Claude Opus 4.5 on November 24, 2025.
1 December 2025DeepSeek released DeepSeek-V3.2 and DeepSeek-V3.2-Speciale on December 1, 2025.
2 December 2025Mistral AI announced and released the Mistral 3 generation on December 2, 2025, including Mistral Large 3 and Ministral 3 models.
11 December 2025OpenAI introduced GPT-5.2 on December 11, 2025.
18 December 2025Google released FunctionGemma in 270M parameter size on December 18, 2025.
18 December 2025Google's official Gemma release ledger records December 18, 2025 as the release of T5Gemma v2 in 270M-270M, 1B-1B, and 4B-4B parameter sizes.
5 February 2026Anthropic introduced Claude Opus 4.6 on February 5, 2026.
16 February 2026Alibaba released Qwen3.5 on February 16, 2026, beginning with the open-sourcing of Qwen3.5-397B-A17B, also named Qwen3.5-Plus, a natively multimodal foundation model.
17 February 2026Anthropic introduced Claude Sonnet 4.6 on February 17, 2026.
19 February 2026Google introduced Gemini 3.1 Pro on February 19, 2026.
3 March 2026OpenAI released GPT-5.3 Instant on March 3, 2026.
5 March 2026OpenAI released GPT-5.4 and GPT-5.4 Pro on March 5, 2026.
16 March 2026Mistral AI announced Mistral Small 4 on March 16, 2026.
23 April 2026OpenAI released GPT-5.5 on April 23, 2026.
24 April 2026DeepSeek introduced DeepSeek-V4 Preview on April 24, 2026.
19 May 2026Google's Gemini API release notes record the generally available release of Gemini 3.5 Flash on May 19, 2026 under the stable model identifier gemini-3.5-flash.
20 May 2026Cohere released Command A+ on May 20, 2026.
22 May 2026Mistral AI made Mistral Medium 3.5 available in public preview on May 22, 2026.
9 July 2026OpenAI launched the GPT-5.6 family for general availability on July 9, 2026.
16 July 2026xAI launched Grok 4.5 on July 16, 2026.
21 July 2026Google records Gemini 3.5 Flash-Lite as generally available on July 21, 2026 under model identifier gemini-3.5-flash-lite.
21 July 2026Google records Gemini 3.6 Flash as generally available on July 21, 2026 under model identifier gemini-3.6-flash.
30 July 2026Google DeepMind introduced Gemini Robotics 2 on July 30, 2026 as a vision-language-action model for intelligent whole-body robot control.
12 August 2026xAI released Grok 4.6 on August 12, 2026.
13 August 2026DeepSeek launched DeepSeek-V4-Pro as generally available on August 13, 2026.
13 August 2026Google records Gemini 3.7 Flash as generally available on August 13, 2026 under model identifier gemini-3.7-flash.
31 August 2026Google's WeatherNext release notes record WeatherNext 3 (v3.0) as launched on August 31, 2026, with live satellite input, 5 km multi-resolution output, hourly initialization.
1 September 2026Anthropic announced Claude Fable 5.1 on September 1, 2026.
2 September 2026Google records Gemini 3.8 Flash with a release date of September 2, 2026 under model identifier gemini-3.8-flash.
3 September 2026OpenAI recorded the release of GPT-6 Astra on September 3, 2026.
3 September 2026Google's Gemini API release notes record September 3, 2026 as the general-availability release of Lyria 3.5.
8 September 2026Google DeepMind introduced AlphaGenome Atlas on September 8, 2026.
10 September 2026DeepSeek released DeepSeek-V4.1-Flash on September 10, 2026.
10 September 2026OpenAI recorded GPT-Live-1 in the API on September 10, 2026.
15 September 2026Google's Gemini API release notes record the general-availability release of Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15.
17 September 2026Google's Gemini API release notes record the release of antigravity-preview-09-2026 on September 17, 2026, replacing and deprecating antigravity-preview-05-2026.
21 September 2026xAI introduced Grok 4.7 on September 21, 2026.
22 September 2026Anthropic introduced Claude Opus 5.5 on September 22, 2026.
22 September 2026Google records Gemini 3.8 Flash TTS as available on September 22, 2026 under model identifier gemini-3.8-flash-tts.
22 September 2026Google recorded Gemini 3.8 Flash-Lite TTS as generally available in the Gemini API on September 22, 2026, under model identifier gemini-3.8-flash-lite-tts.
22 September 2026OpenAI introduced GPT-6 Luna on September 22, 2026.
22 September 2026OpenAI introduced GPT-6 Sol on September 22, 2026.
28 September 2026Anthropic introduced Claude Sonnet 5.5 on September 28, 2026, describing it as the second model in the Claude 5.5 family.
29 September 2026OpenAI's Deployment Safety Hub records the introduction of GPT-6.1 Sol on September 29, 2026.
29 September 2026OpenAI introduced dots on September 29, 2026.
30 September 2026Google DeepMind announced Gemini 4 Argon on September 30, 2026, describing it as a frontier model for complex workflows including software engineering, enterprise knowledge work.
13 October 2026Google published the first-party milestone “Gemini 4 Argon is now available for developers.” on October 13, 2026; the source states an introduction, release, launch, or availability event.
14 October 2026xAI published the first-party milestone “Introducing Grok Voice 3.0” on October 14, 2026; the source states an introduction, release, launch, or availability event.