Skip to content
IFAI

Google DeepMind

Company145 stories

  1. AI modelsReach 52

    Gemini 2.5 Deep Think hits gold medal level at ICPC coding finals

    Google DeepMind just proved its AI can beat almost all elite college coders under tournament conditions. Competing remotely in the 2025 ICPC World Finals, an advanced version of Gemini 2.5 Deep Think solved 10 of 12 problems in 5 hours, ranking second overall and cracking one problem no human team solved. Why it matters: competitive coding requires real reasoning and novel synthesis, proving AI is moving far beyond mere pattern matching.

  2. AI toolsReach 54

    YouTube brings DeepMind's Veo 3 and song remixing to Shorts

    YouTube just unleashed a suite of generative AI tools at its Made on YouTube event. Shorts creators in select regions can now use DeepMind's Veo 3 Fast to generate 480p video clips with audio directly from their phones for free. The platform also added Edit with AI for instant rough cuts and a Lyria 2-powered Speech to Song tool that turns video dialogue into music tracks.

  3. Open sourceReach 45

    Google drops VaultGemma, an LLM built to forget your secrets

    Google and DeepMind just open-sourced VaultGemma, a 1B-parameter model trained from scratch with differential privacy. By mathematically injecting noise into training, it ensures the model never memorizes sensitive sequences. While privacy constraints pull its benchmark performance back to roughly GPT-2 levels, it provides the first empirical scaling laws for provably private AI.

  5. ResearchReach 31

    DeepMind exposes a hard mathematical ceiling in vector search

    Google DeepMind researchers just proved that single-vector embeddings have a fundamental geometric limit that breaks advanced RAG setups. On their new LIMIT stress test, top embedding models from Google and Snowflake saw recall drop below 20%, while old-school BM25 excelled. Why it matters: simply scaling models won't fix retrieval.

  6. AI toolsReach 50

    Google Labs launches Stax to take the guesswork out of LLM evals

    Google Labs and DeepMind just launched Stax, an experimental developer tool built to replace "vibe testing" with structured evaluations. It allows developers to upload test datasets and deploy Gemini-powered pre-built or custom autoraters to score outputs on criteria like factuality, tone, and brand fit.

  7. Generative mediaReach 53

    Google supercharges image editing in Gemini

    Google just integrated a new DeepMind image editing model into the Gemini app that finally keeps character likeness consistent across edits. You can now swap outfits, place yourself anywhere, and blend multiple photos together without your face or pet morphing into someone else. All generated images include invisible SynthID watermarks.

  8. IndustryReach 41

    xAI co-founder Igor Babuschkin departs to launch safety fund

    xAI co-founder Igor Babuschkin just announced he is leaving the Elon Musk-backed startup to launch Babuschkin Ventures, a venture firm focused on AI safety research. Babuschkin previously led engineering at xAI after stints at DeepMind and OpenAI. His exit follows a turbulent stretch of Grok-related controversies over erratic answers and deepfakes.

  9. ResearchReach 32

    DeepMind's Perch AI decodes wild soundscapes

    DeepMind just updated Perch, its open bioacoustics AI model now available on Kaggle. Trained on nearly twice as much data from public sources like Xeno-Canto, the model now handles challenging underwater audio from coral reefs alongside terrestrial birds. Conservationists can search complex audio and build custom wildlife classifiers in under an hour.

  10. AI modelsReach 60

    DeepMind's Genie 3 builds playable worlds on the fly

    Google DeepMind just unveiled Genie 3, a general-purpose world model that generates interactive environments in real time at 24 fps and 720p from simple text prompts. It keeps scenes consistent for minutes and lets users prompt in-world events like weather shifts on the fly. It is currently rolling out as a limited research preview for select academics and creators.

  12. IndustryReach 48

    Meta poaches three top DeepMind researchers

    Meta just hired three key Google DeepMind researchers—Tianhe Yu, Cosmo Du, and Weiyue Wang—who contributed to Google's Gemini models. The move is part of Mark Zuckerberg's aggressive talent push, offering massive compensation packages to lure top researchers from Google, OpenAI, and Anthropic. Meta is going all-in to build out its AGI muscle.

  13. AI modelsReach 53

    Gemini Deep Think scores gold at the International Mathematical Olympiad

    Google DeepMind just announced that an advanced version of Gemini with Deep Think achieved gold-medal performance at IMO 2025. Unlike last year's systems that took days and required translation into formal languages, Gemini worked end-to-end in natural language to solve 5 of 6 problems within the standard 4.5-hour limit. It will soon roll out to Google AI Ultra subscribers.

  14. ResearchReach 35

    OpenAI claims gold-medal math performance on 2025 IMO

    OpenAI just claimed an unreleased experimental reasoning model achieved gold-medal-level performance on the 2025 International Math Olympiad. Tested without web access or extra tools, it solved 5 of 6 problems to score 35 out of 42. Google DeepMind quickly pushed back, arguing no gold claim is valid without official marking guidelines.

  15. AI toolsReach 40

    Google’s AI agent Big Sleep stops a cyberattack before it starts

    Google DeepMind and Project Zero’s AI agent Big Sleep just uncovered a critical SQLite flaw (CVE-2025-6965) before attackers could weaponize it. The team used the agent to isolate the zero-day threat ahead of time. Google claims this marks the first time an AI agent directly thwarted an in-the-wild exploit operation.

  17. Open sourceReach 47

    Google drops open-source MedGemma models for health AI

    Google just released MedGemma 27B Multimodal and a 400M-parameter image encoder called MedSigLIP for health AI development. Both run on a single GPU—with lightweight variants even fitting on mobile devices—to help researchers build private, tailored medical tools without relying on closed APIs.

  20. Generative mediaReach 36

    Google DeepMind's Veo joins Hollywood for Tribeca short film

    Google DeepMind just premiered its first AI-assisted short film, "ANCESTRA," in collaboration with director Darren Aronofsky's Primordial Soup at the Tribeca Festival. The crew paired live-action scenes with the Veo video model, using new motion-matching and object-insertion tools to blend an AI-generated newborn directly into real footage. It marks a big step toward making generative video usable in real Hollywood pipelines.