As of 2020, DeepMind has published over a thousand papers, including thirteen papers that were accepted by
Nature or
Science. DeepMind received media attention during the AlphaGo period; according to a
LexisNexis search, 1842 published news stories mentioned DeepMind in 2016, declining to 1363 in 2019.
Games Unlike earlier AIs, such as
IBM's
Deep Blue or
Watson, which were developed for a pre-defined purpose and only function within that scope, DeepMind's initial algorithms were intended to be general. They used
reinforcement learning, an algorithm that learns from experience using only raw pixels as data input. Their initial approach used
deep Q-learning with a
convolutional neural network. They tested the system on video games, notably early
arcade games, such as
Space Invaders or
Breakout. In July 2018, researchers from DeepMind trained one of its systems to play the computer game
Quake III Arena. In 2013, DeepMind published research on an AI system that surpassed human abilities in games such as
Pong,
Breakout and
Enduro, while surpassing state of the art performance on
Seaquest,
Beamrider, and
Q*bert. This work reportedly led to the company's acquisition by Google. DeepMind's AI had been applied to video games made in the 1970s and
1980s; work was ongoing for more complex 3D games such as
Quake, which first appeared in the 1990s. an AI Agent which surpasses human level performance on all 57 games of the Atari 2600 suite. In July 2022, DeepMind announced the development of DeepNash, a model-free
multi-agent reinforcement learning system capable of playing the board game
Stratego at the level of a human expert.
AlphaGo and successors In October 2015, a
computer Go program called AlphaGo, developed by DeepMind, beat the European Go champion
Fan Hui, a
2 dan (out of 9 dan possible) professional, five to zero. This was the first time an artificial intelligence (AI) defeated a professional Go player. Previously, computers were only known to have played Go at "amateur" level. Go is considered much more difficult for computers to win compared to other games like
chess, due to the much larger number of possibilities, making it prohibitively difficult for traditional AI methods such as
brute-force. In 2017, an improved version,
AlphaGo Zero, defeated AlphaGo in a hundred out of a hundred games. Later that year,
AlphaZero, a modified version of AlphaGo Zero, gained superhuman abilities at chess and shogi. In 2019, DeepMind released a new model named
MuZero that mastered the domains of
Go,
chess,
shogi, and
Atari 2600 games without human data, domain knowledge, or known rules. AlphaGo technology was developed based on
deep reinforcement learning, making it different from the AI technologies then on the market. The data fed into the AlphaGo algorithm consisted of various moves based on historical tournament data. The number of moves was increased gradually until over 30 million of them were processed. The aim was to have the system mimic the human player, as represented by the input data, and eventually become better. It played against itself and learned from the outcomes; thus, it learned to improve itself over the time and increased its winning rate as a result. AlphaGo used two deep neural networks: a policy network to evaluate move probabilities and a value network to assess positions. The policy network trained via supervised learning, and was subsequently refined by policy-gradient
reinforcement learning. The value network learned to predict winners of games played by the policy network against itself. After training, these networks employed a lookahead
Monte Carlo tree search, using the policy network to identify candidate high-probability moves, while the value network (in conjunction with Monte Carlo rollouts using a fast rollout policy) evaluated tree positions. In contrast, AlphaGo Zero was trained without being fed data of human-played games. Instead it generated its own data, playing millions of games against itself. It used a single neural network, rather than separate policy and value networks. Its simplified tree search relied upon this neural network to evaluate positions and sample moves. A new reinforcement learning algorithm incorporated lookahead search inside the training loop. Ultimately, it needed much less computing power than AlphaGo, running on four specialized AI processors (Google
TPUs), instead of AlphaGo's 48. It also required less training time, being able to beat its predecessor after just three days, compared with months required for the original AlphaGo. Similarly, AlphaZero also learned via
self-play. Researchers applied MuZero to solve the real world challenge of video compression with a set number of bits with respect to Internet traffic on sites such as
YouTube,
Twitch, and
Google Meet. The goal of MuZero is to optimally compress the video so the quality of the video is maintained with a reduction in data. The final result using MuZero was a 6.28% average reduction in bitrate.
AlphaStar In 2016, Hassabis discussed the game
StarCraft as a future challenge, since it requires strategic thinking and handling imperfect information. In January 2019, DeepMind introduced AlphaStar, a program playing the real-time strategy game
StarCraft II. AlphaStar used reinforcement learning based on replays from human players, and then played against itself to enhance its skills. At the time of the presentation, AlphaStar had knowledge equivalent to 200 years of playing time. It won 10 consecutive matches against two professional players, although it had the unfair advantage of being able to see the entire field, unlike a human player who has to move the camera manually. A preliminary version in which that advantage was fixed lost a subsequent match. In July 2019, AlphaStar began playing against random humans on the public 1v1 European multiplayer ladder. Unlike the first iteration of AlphaStar, which played only
Protoss v. Protoss, this one played as all of the game's races, and had earlier unfair advantages fixed. By October 2019, AlphaStar had reached Grandmaster level on the
StarCraft II ladder on all three
StarCraft races, becoming the first AI to reach the top league of a widely popular
esport without any game restrictions.
Datacenter operation In 2014, a datacenter engineer at Google began using supervised machine learning to predict
power usage effectiveness (PUE) of datacenters at Google. The system was deployed in production to allow operators to simulate control strategies and pick the one that saves the most energy. In 2016, inspired by AlphaGo, he contacted DeepMind to apply
reinforcement learning (RL) to train a system that could also recommend actions. It was tested on a live datacenter. The system read from sensor readings and recommended actions to take, and human engineers would implement the actions. Though the human engineers found its recommendations unintuitive, they satisfied all safety constraints, and led to a 15% saving in PUE. The system was deployed more widely across Google, with datacenter controllers receiving email recommendations from the system every 15 minutes. Eventually a more mature and more autonomous system was deployed, where the AI's actions are checked against safety constraints and implemented autonomously if verified safe, and human operators would supervise the AI and may override. The system led to a 30% saving in PUE. The system produced cooling strategies that surprised long-time operators, such as exploiting winter conditions to produce colder than normal water. Google subsequently collaborated with
Trane Technologies to deploy similar RL-based systems on
HVAC of facilities outside of Google.
Protein folding In 2016, DeepMind turned its artificial intelligence to
protein folding, a long-standing problem in
molecular biology. In December 2018, DeepMind's AlphaFold won the 13th
Critical Assessment of Techniques for Protein Structure Prediction (CASP) by successfully predicting the most accurate structure for 25 out of 43 proteins. "This is a lighthouse project, our first major investment in terms of people and resources into a fundamental, very important, real-world scientific problem," Hassabis said to
The Guardian. In 2020, in the 14th CASP, AlphaFold's predictions achieved an accuracy score regarded as comparable with lab techniques. Andriy Kryshtafovych, one of the panel of scientific adjudicators, described the achievement as "truly remarkable", and said the problem of predicting how proteins fold had been "largely solved". In July 2021, the open-source RoseTTAFold and AlphaFold2 were released to allow scientists to run their own versions of the tools. A week later DeepMind announced that AlphaFold had completed its prediction of nearly all human proteins as well as the entire
proteomes of 20 other widely studied organisms. The structures were released on the AlphaFold Protein Structure Database. In July 2022, it was announced that the predictions of over 200 million proteins, representing virtually all known proteins, would be released on the AlphaFold database. In October 2024, Hassabis and
John Jumper received half of the 2024
Nobel Prize in Chemistry jointly for protein structure prediction, citing AlphaFold2 achievement.
Language models In 2016, DeepMind introduced
WaveNet, a
text-to-speech system. It was originally too computationally intensive for use in consumer products, but in late 2017 it became ready for use in consumer applications such as
Google Assistant. In 2018 Google launched a commercial text-to-speech product, Cloud Text-to-Speech, based on WaveNet. In 2018, DeepMind introduced a more efficient model called WaveRNN co-developed with
Google AI. In 2020 WaveNetEQ, a packet loss concealment method based on a WaveRNN architecture, was presented. In 2019, Google started to roll WaveRNN with WavenetEQ out to
Google Duo users. Released in May 2022,
Gato is a polyvalent
multimodal model. It was trained on 604 tasks, such as image captioning, dialogue, or stacking blocks. On 450 of these tasks, Gato outperformed human experts at least half of the time, according to DeepMind. Unlike models like MuZero, Gato does not need to be retrained to switch from one task to the other.
Sparrow is an artificial intelligence-powered
chatbot developed by DeepMind to build safer machine learning systems by using a mix of human feedback and Google search suggestions.
Chinchilla is a language model developed by DeepMind. DeepMind posted a blog post on 28 April 2022 on a single visual language model (VLM) named Flamingo that can accurately describe a picture of something with just a few training images.
AlphaCode In 2022, DeepMind unveiled AlphaCode, an
AI-powered coding engine that creates
computer programs at a rate comparable to that of an average programmer, with the company testing the system against coding challenges created by
Codeforces utilized in human
competitive programming competitions. AlphaCode earned a rank equivalent to 54% of the median score on Codeforces after being trained on
GitHub data and Codeforce problems and solutions. The program was required to come up with a unique solution and stopped from duplicating answers.
Gemini Gemini is a
multimodal large language model which was released on 6 December 2023. It is the successor of Google's
LaMDA and
PaLM 2 language models and sought to challenge OpenAI's
GPT-4. Gemini comes in 3 sizes: Nano, Pro, and Ultra. Gemini is also the name of the chatbot that integrates Gemini (and which was previously called
Bard). On 12 December 2024, Google released Gemini 2.0 Flash, the first model in the Gemini 2.0 series. It notably features expanded multimodality, with the ability to also generate images and audio, and is part of Google's broader plans to integrate advanced AI into
autonomous agents. On 25 March 2025, Google released Gemini 2.5, a reasoning model that stops to "think" before giving a response. Google announced that all future models will also have reasoning ability. On 30 March 2025, Google released Gemini 2.5 to all free users. On 18 November 2025, Google released Gemini 3 Pro, a reasoning model which is fully multimodal. It was fully integrated with Google Search and AI Mode the same day.
Gemma Gemma is a collection of open-weight large language models. The first ones were released on 21 February 2024 and are available in two distinct sizes: a 7 billion parameter model optimized for GPU and TPU usage, and a 2 billion parameter model designed for CPU and on-device applications. Gemma models were trained on up to 6 trillion tokens of text, employing similar architectures, datasets, and training methodologies as the Gemini model set. In June 2024, Google started releasing Gemma 2 models. In December 2024, Google introduced
PaliGemma 2, an upgraded vision-language model. In February 2025, they launched
PaliGemma 2 Mix, a version fine-tuned for multiple tasks. It is available in 3B, 10B, and 28B parameters with 224px and 448px resolutions. In March 2025, Google released Gemma 3, calling it the most capable model that can be run on a single GPU. It has four available sizes: 1B, 4B, 12B, and 27B. In March 2025, Google introduced TxGemma, an open-source model designed to improve the efficiency of therapeutics development. In April 2025, Google introduced DolphinGemma, a research artificial intelligence model designed to hopefully decode dolphin communication. They want to train a foundation model that can learn the structure of dolphin vocalizations and generate novel dolphin-like sound sequences.
SIMA In March 2024, DeepMind introduced Scalable Instructable Multiword Agent, or SIMA, an AI agent capable of understanding and following natural language instructions to complete tasks across various 3D virtual environments. Trained on nine video games from eight studios and four research environments, SIMA demonstrated adaptability to new tasks and settings without requiring access to game source code or APIs. The agent comprises pre-trained computer vision and language models fine-tuned on gaming data, with language being crucial for understanding and completing given tasks as instructed. DeepMind's research aimed to develop more helpful AI agents by translating advanced AI capabilities into real-world actions through a language interface.
Habermas machine In 2024, Google Deepmind published the results of an experiment where they trained two
large language models to help identify and present areas of overlap among a few thousand group members they had recruited online using techniques like
sortition to get a representative sample of participants. The project is named in honor of
Jürgen Habermas. In one experiment, the participants rated the summaries by the AI higher than the human moderator 56% of the time. In December 2024,
Google released Veo 2, available via VideoFX. It supports
4K resolution video generation, and has an improved understanding of physics. In April 2025, Google announced that Veo 2 became available for advanced users on Gemini App. In May 2025, Google released Veo 3, which not only generates videos but also creates synchronized audio — including dialogue, sound effects, and ambient noise — to match the visuals. Google also announced Flow, a video-creation tool powered by Veo and
Imagen.
Music generation Google DeepMind developed Lyria, a text-to-music model. As of August 2025, it is available on Vertex AI and the Gemini API. On February 18, 2026, DeepMind released Lyria 3. On March 25, 2026, DeepMind released Lyria 3 Pro which allows users to create longer tracks with more structural awareness.
Environment generation In March 2024, DeepMind introduced "
Genie" (Generative Interactive Environments), an AI model that can generate game-like, action-controllable virtual worlds based on textual descriptions, images, or sketches. Built as an autoregressive
latent diffusion model, Genie enables frame-by-frame interactivity without requiring labeled action data for training. Its successor, Genie 2, released in December 2024, expanded these capabilities to generate diverse and interactive 3D environments. Genie 3 was released in August 2025, with higher-resolution world generations and multiple minutes of visual consistency. On January 29, 2026, DeepMind released Project Genie to AI Ultra subscribers.
Robotics Released in June 2023, RoboCat is an AI model that can control robotic arms. The model can adapt to new models of robotic arms, and to new types of tasks. In March 2025, DeepMind launched two AI models, Gemini Robotics and Gemini Robotics-ER, aimed at improving how robots interact with the physical world and released Gemini Robotics 1.5 in September 2025. In April 2026, DeepMind launched Gemini Robotics ER-1.6, which is an upgrade to their previous model.
Others Football DeepMind researchers have applied machine learning models to the sport of
football, often referred to as soccer in North America, modelling the behaviour of football players, including the goalkeeper, defenders, and strikers during different scenarios such as penalty kicks. The researchers used heat maps and cluster analysis to organize players based on their tendency to behave a certain way during the game when confronted with a decision on how to score or prevent the other team from scoring. The researchers mention that machine learning models could be used to democratize the football industry by automatically selecting interesting video clips of the game that serve as highlights. This can be done by searching videos for certain events, which is possible because video analysis is an established field of machine learning. This is also possible because of extensive sports analytics based on data including annotated passes or shots, sensors that capture data about the players movements many times over the course of a game, and game theory models.
Archaeology Google has unveiled a new archaeology document program, named Ithaca after
the Greek island in Homer's
Odyssey. This deep neural network helps researchers restore the empty text of damaged Greek documents, and to identify their date and geographical origin. The work builds on another text analysis network that DeepMind released in 2019, named Pythia. However, according to
Anthony Cheetham, GNoME did not make "a useful, practical contribution to the experimental materials scientists." A review article by Cheetham and Ram Seshadri were unable to identify any "strikingly novel" materials found by GNoME, with most being minor variants of already-known materials.
Mathematics AlphaTensor In October 2022, DeepMind released
AlphaTensor, which used reinforcement learning techniques similar to those in AlphaGo, to find novel
algorithms for matrix multiplication. In the special case of multiplying two 4×4 matrices with
integer entries, where only the evenness or oddness of the entries is recorded, AlphaTensor found an algorithm requiring only 47 distinct multiplications; the previous optimum, known since 1969, was the more general
Strassen algorithm, using 49 multiplications. Computer scientist Josh Alman described AlphaTensor as "a proof of concept for something that could become a breakthrough", while
Vassilevska Williams called it "a little overhyped" Traditional geometry programs are
symbolic engines that rely exclusively on human-coded
rules to generate rigorous proofs, which makes them lack flexibility in unusual situations. AlphaGeometry combines such a symbolic engine with a specialized
large language model trained on
synthetic data of geometrical proofs. When the symbolic engine doesn't manage to find a formal and rigorous proof on its own, it solicits the large language model, which suggests a geometrical construct to move forward. However, it is unclear how applicable this method is to other domains of mathematics or reasoning, because symbolic engines rely on domain-specific rules and because of the need for synthetic data.
AlphaDev In June 2023, Deepmind announced that
AlphaDev, which searches for improved computer science algorithms using
reinforcement learning, discovered a more efficient way of coding a sorting algorithm and a hashing algorithm. The new sorting algorithm was 70% faster for shorter sequences and 1.7% faster for sequences exceeding 250,000 elements, and the new hashing algorithm was 30% faster in some cases. The sorting algorithm was accepted into the
C++ Standard Library sorting algorithms, and was the first change to those algorithms in more than a decade and the first update to involve an algorithm discovered using AI. The hashing algorithm was released to an opensource library. Google estimates that these two algorithms are used trillions of times every day.
AlphaEvolve In May 2025, Google DeepMind unveiled
AlphaEvolve, an
evolutionary coding agent using LLMs like Gemini to design optimized algorithms. AlphaEvolve begins each optimization process with an initial algorithm and metrics to evaluate the quality of a solution. At each step, it uses the LLM to generate variations of the algorithms or combine them, and selects the best candidates for further iterations. AlphaEvolve has made several algorithmic discoveries, including in matrix multiplication. According to Google, when tested on 50 open
mathematical problems, AlphaEvolve was able to match the efficiency of state-of-the-art algorithms in 75% of cases, and discovered improved solutions 20% of the time, such as with the
kissing number problem in 11 dimensions. It also developed a new heuristic for data centre scheduling, recovering on average 0.7% of Google's worldwide compute resources. Multiple independent researchers remained unconvinced, citing a lack of direct public benchmarks and independent proof of its claimed superiority over existing commercial chip design tools. The TPU chips were co-designed with
Broadcom.
Communications of the ACM noted that despite substantial publicity, DeepMind had not provided the comparative benchmarks long requested by experts, leaving some skepticism in the field. Similarly,
New Scientist reported that while Google claims AlphaChip has produced “superhuman” chip layouts now used in production, external specialists called for transparent performance data to substantiate these assertions and enable fair comparisons with current state-of-the-art methods.
Safety Google Research released a paper in 2016 regarding
AI safety and avoiding undesirable behaviour during the AI learning process. In 2017 DeepMind released GridWorld, an open-source testbed for evaluating whether an algorithm learns to disable its
kill switch or otherwise exhibits certain undesirable behaviours. The Robot Constitution is a security ruleset part of AutoRT set by DeepMind in January 2024 for its AI products. The rules are inspired by
Asimov's
Three Laws of Robotics. The rules are applied to the underlying
large language models of the helper robots. Rule number 1 is a robot “may not injure a human being”.
Weather prediction Google DeepMind developed an AI-based weather prediction system called Weather Lab, which significantly improved tropical cyclone forecasting. Launched in mid-2025, this model utilized stochastic neural networks trained on 45 years of global weather and cyclone data, enabling it to predict cyclone formation, track, intensity, and structure with multiple probabilistic forecasts up to 15 days in advance. During the 2025 Atlantic hurricane season, DeepMind's Weather Lab outperformed traditional physics-based models, including the
US National Weather Service's Global Forecast System, in both track and intensity predictions, earning notable recognition from meteorologists and aiding hurricane forecasting efforts by the
US National Hurricane Center. This marked a substantial advancement in weather modeling, demonstrating the potential for AI to enhance the speed and accuracy of severe weather forecasts.
Miscellaneous contributions to Google DeepMind (alongside other Alphabet AI researchers) assists
Google Play's personalized app recommendations. == DeepMind Health ==