Gemini 3

For centuries, the scientific method has been our best tool for progress. But today, there’s so much data out there that it’s impossible for any one researcher to connect all the dots. We want to fix

For centuries, the scientific method has been our best tool for progress. But today, there’s so much data out there that it’s impossible for any one researcher to connect all the dots. We want to fix that: Introducing Gemini for Science, a collection of science tools and https://t.co/knRWV2JJsR

We partnered with artists, designers, and builders to create new AI tools that solve real problems in their creative workflows. Here’s what’s new: — Introducing Google Pics in @GoogleWorkspace: A bran

We partnered with artists, designers, and builders to create new AI tools that solve real problems in their creative workflows. Here’s what’s new: — Introducing Google Pics in @GoogleWorkspace: A brand-new image creation & editing tool. Move and resize objects, add text, and https://t.co/e5nJrAfUHP

Build your next story with Gemini Omni. https://t.co/0Gd5Nlw26U

New upgrades to the @GeminiApp are you helping you get more done: ✨Gemini Spark is your 24/7 personal AI agent that can take action on your behalf, under your direction. It seamlessly integrates with

New upgrades to the @GeminiApp are you helping you get more done: ✨Gemini Spark is your 24/7 personal AI agent that can take action on your behalf, under your direction. It seamlessly integrates with @Gmail, @GoogleDocs, and Slides to automate your workflows and, best of all, https://t.co/pMCS05HAhB

A few weeks ago, we asked our community to use @GoogleAIStudio or Canvas in @GeminiApp to help us create the Google I/O countdown. Thanks SO much to everyone who submitted, and special shoutout to the

A few weeks ago, we asked our community to use @GoogleAIStudio or Canvas in @GeminiApp to help us create the Google I/O countdown. Thanks SO much to everyone who submitted, and special shoutout to the creators whose submissions helped us set the right ~vibes~ on the stage today: https://t.co/A1zMExmEVM

Some fun Gemini Omni use cases from the community👇🧵 (We’ll keep updating this thread throughout the day)

We were able to sit down with the @GoogleDeepmind team behind the new Gemini Omni Flash model to hear all of their behind-the-scenes stories, memorable moments, and many, many (occasionally embarrassi

We were able to sit down with the @GoogleDeepmind team behind the new Gemini Omni Flash model to hear all of their behind-the-scenes stories, memorable moments, and many, many (occasionally embarrassing) video generations. Watch the full Release Notes episode here: https://t.co/cA911hq2IL

By now, you've probably heard about Gemini Omni, our new model designed to create anything from any input, starting with video. But... what's the big deal? Let’s break it down 🧵👇 https://t.co/QbxMNZ

By now, you've probably heard about Gemini Omni, our new model designed to create anything from any input, starting with video. But... what's the big deal? Let’s break it down 🧵👇 https://t.co/QbxMNZa2Wx

We want to help scientists discover their next breakthrough with AI. Gemini for Science is our new suite of experimental tools to help them explore more hypotheses, validate work at scale, unpack lite

We want to help scientists discover their next breakthrough with AI. Gemini for Science is our new suite of experimental tools to help them explore more hypotheses, validate work at scale, unpack literature with ease, and more 🧵 https://t.co/RyHvlZCS7u

Last week, we launched Gemini 3.1 TTS, our latest and best text-to-speech model. This new model introduces [awe] audio tags, an intuitive way to guide vocal style, pace, and delivery. Here are some ti

Last week, we launched Gemini 3.1 TTS, our latest and best text-to-speech model. This new model introduces [awe] audio tags, an intuitive way to guide vocal style, pace, and delivery. Here are some tips on the best ways to use audio tags in your prompts: 1. All inline tags must https://t.co/YDbBLs5Dcp

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devic

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: https://t.co/JgwRZvQHgF

Watch how fast Gemini 3.1 Flash-Lite can generate websites. ⚡ This browser creates each page in real-time as you click, search, and navigate. Give it a try → https://t.co/iibqowLwme https://t.co/h1RJ8

Watch how fast Gemini 3.1 Flash-Lite can generate websites. ⚡ This browser creates each page in real-time as you click, search, and navigate. Give it a try → https://t.co/iibqowLwme https://t.co/h1RJ86cB54

ICYMI, here’s a recap of this week’s launches: — Gemini 3.1 Flash-Lite (in preview), our most cost-efficient Gemini 3 series model yet — Cinematic Video Overviews from @NotebookLM, turning your source

ICYMI, here’s a recap of this week’s launches: — Gemini 3.1 Flash-Lite (in preview), our most cost-efficient Gemini 3 series model yet — Cinematic Video Overviews from @NotebookLM, turning your sources into bespoke, immersive videos — 10 custom styles for NotebookLM

Here are a couple examples of how Gemini 3.1 Flash-Lite can solve real-world problems: First, this high-volume image sorter showcases the model’s ability to quickly analyze and sort large amounts of c

Here are a couple examples of how Gemini 3.1 Flash-Lite can solve real-world problems: First, this high-volume image sorter showcases the model’s ability to quickly analyze and sort large amounts of content, like pictures (something that could have been too expensive or slow in https://t.co/sl8ajXKHmk

Smarter. Faster. Gemini 3.1 Flash-Lite is here⚡ The model offers uncompromising speed & intelligence at scale by focusing on: — Cost-efficiency: Priced at just $0.25/1M input and $1.50/1M output token

Smarter. Faster. Gemini 3.1 Flash-Lite is here⚡ The model offers uncompromising speed & intelligence at scale by focusing on: — Cost-efficiency: Priced at just $0.25/1M input and $1.50/1M output tokens, it gets work done faster at a fraction of the cost of larger models, https://t.co/icrk62FTJ3

MERRIN: A Benchmark for Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments

Motivated by the underspecified, multi-hop nature of search queries and the multimodal, heterogeneous, and often conflicting nature of real-world web results, we introduce MERRIN (Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments), a human-annotated benchmark for evaluating search-augmented agents. MERRIN measures AI agents' ability to identify relevant modalities, retrieve multimodal evidence, and perform multi-hop reasoning over noisy web sources. It differs from prior work in three important aspects: (1) using natural language queries without explicit modality cues, (2) incorporating underexplored modalities such as video and audio, and (3) requiring the retrieval of complex, often noisy or conflicting multimodal evidence during web search. We evaluate diverse search agents powered by ten models, including strong closed-source models (e.g., GPT-5.4-mini, Gemini 3/3.1 Flash/Pro) and open-weight models (Qwen3-4B/30B/235B), across three search settings (no search, native search, and agentic search). Our results show that MERRIN is highly challenging: the average accuracy across all agents is 22.3%, with the best-performing agent reaching only 40.1%. We further observe that while stronger agents like Gemini Deep Research achieve higher performance, gains are modest due to over-exploration; they take more steps and use more tools, but are often distracted by conflicting or partially relevant web content, leading to incorrect answers. Compared to humans, these agents consume more resources yet achieve lower accuracy, largely due to inefficient source selection and an overreliance on text modalities. These findings highlight the need for search agents capable of robust search and reasoning across diverse modalities in noisy web environments, making MERRIN a valuable testbed for evaluating such capabilities.

OmniScript: Towards Audio-Visual Script Generation for Long-Form Cinematic Video

Current multimodal large language models (MLLMs) have demonstrated remarkable capabilities in short-form video understanding, yet translating long-form cinematic videos into detailed, temporally grounded scripts remains a significant challenge. This paper introduces the novel video-to-script (V2S) task, aiming to generate hierarchical, scene-by-scene scripts encompassing character actions, dialogues, expressions, and audio cues. To facilitate this, we construct a first-of-its-kind human-annotated benchmark and propose a temporally-aware hierarchical evaluation framework. Furthermore, we present OmniScript, an 8B-parameter omni-modal (audio-visual) language model tailored for long-form narrative comprehension. OmniScript is trained via a progressive pipeline that leverages chain-of-thought supervised fine-tuning for plot and character reasoning, followed by reinforcement learning using temporally segmented rewards. Extensive experiments demonstrate that despite its parameter efficiency, OmniScript significantly outperforms larger open-source models and achieves performance comparable to state-of-the-art proprietary models, including Gemini 3-Pro, in both temporal localization and multi-field semantic accuracy.

HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents

We introduce HY-Embodied-0.5, a family of foundation models specifically designed for real-world embodied agents. To bridge the gap between general Vision-Language Models (VLMs) and the demands of embodied agents, our models are developed to enhance the core capabilities required by embodied intelligence: spatial and temporal visual perception, alongside advanced embodied reasoning for prediction, interaction, and planning. The HY-Embodied-0.5 suite comprises two primary variants: an efficient model with 2B activated parameters designed for edge deployment, and a powerful model with 32B activated parameters targeted for complex reasoning. To support the fine-grained visual perception essential for embodied tasks, we adopt a Mixture-of-Transformers (MoT) architecture to enable modality-specific computing. By incorporating latent tokens, this design effectively enhances the perceptual representation of the models. To improve reasoning capabilities, we introduce an iterative, self-evolving post-training paradigm. Furthermore, we employ on-policy distillation to transfer the advanced capabilities of the large model to the smaller variant, thereby maximizing the performance potential of the compact model. Extensive evaluations across 22 benchmarks, spanning visual perception, spatial reasoning, and embodied understanding, demonstrate the effectiveness of our approach. Our MoT-2B model outperforms similarly sized state-of-the-art models on 16 benchmarks, while the 32B variant achieves performance comparable to frontier models such as Gemini 3.0 Pro. In downstream robot control experiments, we leverage our robust VLM foundation to train an effective Vision-Language-Action (VLA) model, achieving compelling results in real-world physical evaluations. Code and models are open-sourced at https://github.com/Tencent-Hunyuan/HY-Embodied.

Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization

We present Kernel-Smith, a framework for high-performance GPU kernel and operator generation that combines a stable evaluation-driven evolutionary agent with an evolution-oriented post-training recipe. On the agent side, Kernel-Smith maintains a population of executable candidates and iteratively improves them using an archive of top-performing and diverse programs together with structured execution feedback on compilation, correctness, and speedup. To make this search reliable, we build backend-specific evaluation services for Triton on NVIDIA GPUs and Maca on MetaX GPUs. On the training side, we convert long-horizon evolution trajectories into step-centric supervision and reinforcement learning signals by retaining correctness-preserving, high-gain revisions, so that the model is optimized as a strong local improver inside the evolutionary loop rather than as a one-shot generator. Under a unified evolutionary protocol, Kernel-Smith-235B-RL achieves state-of-the-art overall performance on KernelBench with Nvidia Triton backend, attaining the best average speedup ratio and outperforming frontier proprietary models including Gemini-3.0-pro and Claude-4.6-opus. We further validate the framework on the MetaX MACA backend, where our Kernel-Smith-MACA-30B surpasses large-scale counterparts such as DeepSeek-V3.2-think and Qwen3-235B-2507-think, highlighting potential for seamless adaptation across heterogeneous platforms. Beyond benchmark results, the same workflow produces upstream contributions to production systems including SGLang and LMDeploy, demonstrating that LLM-driven kernel optimization can transfer from controlled evaluation to practical deployment.

$R$-equivalence on Cubic Surfaces I: Existing Cases with Non-Trivial Universal Equivalence

Let $V$ be a smooth cubic surface over a $p$-adic field $k$ with good reduction. Swinnerton-Dyer (1981) proved that $R$-equivalence is trivial on $V(k)$ except perhaps if $V$ is one of three special types--those whose $R$-equivalence he could not bound by proving the universal (admissible) equivalence is trivial. We consider all surfaces $V$ currently known to have non-trivial universal equivalence. Beyond being intractable to Swinnerton-Dyer's approach, we observe that if these surfaces also had non-trivial $R$-equivalence, they would contradict Colliot-Thélène and Sansuc's conjecture regarding the $k$-rationality of universal torsors for geometrically rational surfaces. By devising new methods to study $R$-equivalence, we prove that for 2-adic surfaces with all-Eckardt reductions (the third special type, which contains every existing case of non-trivial universal equivalence), $R$-equivalence is trivial or of exponent 2. For the explicit cases, we confirm triviality: the diagonal cubic $X^3+Y^3+Z^3+ζ_3 T^3=0$ over $\mathbb{Q}_2(ζ_3)$--answering a long-standing question of Manin's (Cubic Forms, 1972)--and the cubic with universal equivalence of exponent 2 (Kanevsky, 1982). This is the first in a series of works derived from a year of interactions with generative AI models such as AlphaEvolve and Gemini 3 Deep Think, with the latter proving many of our lemmas. We disclose the timeline and nature of their use towards this paper, and describe our broader AI-assisted research program in a companion report (in preparation).

Differential Harm Propensity in Personalized LLM Agents: The Curious Case of Mental Health Disclosure

Large language models (LLMs) are increasingly deployed as tool-using agents, shifting safety concerns from harmful text generation to harmful task completion. Deployed systems often condition on user profiles or persistent memory, yet agent safety evaluations typically ignore personalization signals. To address this gap, we investigated how mental health disclosure, a sensitive and realistic user-context cue, affects harmful behavior in agentic settings. Building on the AgentHarm benchmark, we evaluated frontier and open-source LLMs on multi-step malicious tasks (and their benign counterparts) under controlled prompt conditions that vary user-context personalization (no bio, bio-only, bio+mental health disclosure) and include a lightweight jailbreak injection. Our results reveal that harmful task completion is non-trivial across models: frontier lab models (e.g., GPT 5.2, Claude Sonnet 4.5, Gemini 3-Pro) still complete a measurable fraction of harmful tasks, while an open model (DeepSeek 3.2) exhibits substantially higher harmful completion. Adding a bio-only context generally reduces harm scores and increases refusals. Adding an explicit mental health disclosure often shifts outcomes further in the same direction, though effects are modest and not uniformly reliable after multiple-testing correction. Importantly, the refusal increase also appears on benign tasks, indicating a safety--utility trade-off via over-refusal. Finally, jailbreak prompting sharply elevates harm relative to benign conditions and can weaken or override the protective shift induced by personalization. Taken together, our results indicate that personalization can act as a weak protective factor in agentic misuse settings, but it is fragile under minimal adversarial pressure, highlighting the need for personalization-aware evaluations and safeguards that remain robust across user-context conditions.

VTC-Bench: Evaluating Agentic Multimodal Models via Compositional Visual Tool Chaining

Recent advancements extend Multimodal Large Language Models (MLLMs) beyond standard visual question answering to utilizing external tools for advanced visual tasks. Despite this progress, precisely executing and effectively composing diverse tools for complex tasks remain persistent bottleneck. Constrained by sparse tool-sets and simple tool-use trajectories, existing benchmarks fail to capture complex and diverse tool interactions, falling short in evaluating model performance under practical, real-world conditions. To bridge this gap, we introduce VisualToolChain-Bench(VTC-Bench), a comprehensive benchmark designed to evaluate tool-use proficiency in MLLMs. To align with realistic computer vision pipelines, our framework features 32 diverse OpenCV-based visual operations. This rich tool-set enables extensive combinations, allowing VTC-Bench to rigorously assess multi-tool composition and long-horizon, multi-step plan execution. For precise evaluation, we provide 680 curated problems structured across a nine-category cognitive hierarchy, each with ground-truth execution trajectories. Extensive experiments on 19 leading MLLMs reveal critical limitations in current models' visual agentic capabilities. Specifically, models struggle to adapt to diverse tool-sets and generalize to unseen operations, with the leading model Gemini-3.0-Pro only achieving 51% on our benchmark. Furthermore, multi-tool composition remains a persistent challenge. When facing complex tasks, models struggle to formulate efficient execution plans, relying heavily on a narrow, suboptimal subset of familiar functions rather than selecting the optimal tools. By identifying these fundamental challenges, VTC-Bench establishes a rigorous baseline to guide the development of more generalized visual agentic models.

AI Can Learn Scientific Taste

Great scientists have strong judgement and foresight, closely tied to what we call scientific taste. Here, we use the term to refer to the capacity to judge and propose research ideas with high potential impact. However, most relative research focuses on improving an AI scientist's executive capability, while enhancing an AI's scientific taste remains underexplored. In this work, we propose Reinforcement Learning from Community Feedback (RLCF), a training paradigm that uses large-scale community signals as supervision, and formulate scientific taste learning as a preference modeling and alignment problem. For preference modeling, we train Scientific Judge on 700K field- and time-matched pairs of high- vs. low-citation papers to judge ideas. For preference alignment, using Scientific Judge as a reward model, we train a policy model, Scientific Thinker, to propose research ideas with high potential impact. Experiments show Scientific Judge outperforms SOTA LLMs (e.g., GPT-5.2, Gemini 3 Pro) and generalizes to future-year test, unseen fields, and peer-review preference. Furthermore, Scientific Thinker proposes research ideas with higher potential impact than baselines. Our findings show that AI can learn scientific taste, marking a key step toward reaching human-level AI scientists.

Qianfan-OCR: A Unified End-to-End Model for Document Intelligence

We present Qianfan-OCR, a 4B-parameter end-to-end vision-language model that unifies document parsing, layout analysis, and document understanding within a single architecture. It performs direct image-to-Markdown conversion and supports diverse prompt-driven tasks including table extraction, chart understanding, document QA, and key information extraction. To address the loss of explicit layout analysis in end-to-end OCR, we propose Layout-as-Thought, an optional thinking phase triggered by special think tokens that generates structured layout representations -- bounding boxes, element types, and reading order -- before producing final outputs, recovering layout grounding capabilities while improving accuracy on complex layouts. Qianfan-OCR ranks first among end-to-end models on OmniDocBench v1.5 (93.12) and OlmOCR Bench (79.8), achieves competitive results on OCRBench, CCOCR, DocVQA, and ChartQA against general VLMs of comparable scale, and attains the highest average score on public key information extraction benchmarks, surpassing Gemini-3.1-Pro, Seed-2.0, and Qwen3-VL-235B. The model is publicly accessible via the Baidu AI Cloud Qianfan platform.

gemini-3-flash-preview - SWE-Bench Verified

SWE-Bench Verified resolved rate 75.8

Gemini 3.1 Flash Live: Making audio AI more natural and reliable

Developers can use it to build voice agents that handle complex tasks more reliably. 3.1 Flash Live is available across Google products: For developers in preview via the Gemini Live API in Google AI Studio For enterprises in Gemini Enterprise for Customer Experience For everyone via Search Live and Gemini Live For developers: Robust reasoning and task execution We’ve improved 3.1 Flash Live’s overall quality, making it more reliable for developers and enterprises to build vo

Gemini-3-flash-preview - GAIA

GAIA score 41.5 from gemini-3-flash-with-search